Ying's Opinion

What a bad luck

2007-08-24T20:37:00.000-04:00

This year, my Informs presentation is arranged at Wednesday again. I was looking forward to presenting my transshipment stuff in front of more people than the last year.

Two years in a row to present at Wednesday. What a bad luck! Considering that it is in Seattle, a lot of people living in the east coast would need to take flight in the morning so that they can get home before midnight. So if they don't want to cancel or miss two classes, they won't be able to make any Wednesday session.

Informs need to consider to shift their session starting Sat. instead of Sun. so that there is more people can attend all the sessions. Or they need to think out a way to enlarge the number of parallel sessions. Sun. to Wed. is not a good idea.

Stuck at Chicago

2007-06-29T06:29:00.000-04:00

First trip, I got my bagage at second day after I arrived Chicago and got fever shortly.
Second trip, I got stomach flu at last day at Chicago.
This time, I am stuck at Chicago for two days due to bad weather.
Maybe Chicago knows I hate it and want to keep me as long as possible. Anyway, Chicago is beautiful right now and it is so easy to find a good restaurant with reasonable price.

Back to China for three weeks

2007-06-03T19:37:00.000-04:00

I haven't been back to China for two years. Now it is time. I will departure on June 7th and be back on June 28th.

During those three weeks, I am also going to attend MSOM conference at Tsinghua. It is said that there is more people showing up in each session compared with Informs annual meeting. Let's see whether I can absorb more stuff during the meeting.

Also it is a good chance to talk with my old friends there. When I am in USA, I have very little connection to the industry. It is a big disadvantage when doing applied model for real business environment. One of my little sisters already opened a chain store. I hope to see how it operates daily, where my knowledge can help her, and whethter I can come up some new models.

When I am back to USA, I need to work on my proposal. Ideally, I propose in the middle of August. Though time is scarce resource to me, I will enjoy a short break

Open source codes of each paper

2007-05-20T12:42:00.000-04:00

Usually when you read a journal paper and happen to to like it. So you want to implement their idea. Now the problem occurs, how can I implement those ideas? I need to work out every detail to work it out. Also only when you work it out in detail, you can say you understand it since you make computer understand it.

Why do all papers have their code attached? I will do it for all my papers from now on.

to make my code more organized
to make my work better known by others if they can see how concrete results come out
to improve the work if I can get the feedback from others.

Back to Blog again

2007-05-16T23:00:00.000-04:00

The first rule of Newton principle is that when you stop, you will stop forever unless there is a change of outside force. It can be applied to write blog. Once you stop, you will stop forever. But finally some force disappear. The spring semester is over.

Now I am rusty as a researcher. Too many things distract me. But I will be ready soon.

Lehigh IE rank drops 2 in 2007

2007-04-08T12:38:00.000-04:00

Based on the latest US News, the rank of Lehigh IE drops to 15 from 13 this year. USC and NCSU go above Lehigh this year.

We need more professors!!

What does research do on airline industry

2007-04-02T13:32:00.000-04:00

2007 airline quality rating comes out today. There are two major factors, I guess research field care about. One is on time performance which is related to scheduling problem. The other is deny rate which is related to RM.
http://www.eyewitnessnewstv.com/Global/story.asp?S=6312644
On time performance becomes worse

On-time performance, the report said, worsened last year, with 75.5 percent of
flights arriving on time, compared with 77.3 percent in 2005. JetBlue Airways
took a hit in February, when passengers on 10 planes spent from five hours to 10
1/2 hours sitting on runways at John F. Kennedy International Airport in New
York because of icy weather and gate congestion. It took days for the airline to
recover from the February storm and resume normal operations. It led JetBlue to
establish a customer bill of rights promising vouchers to passengers who
experience delays.

Deny rate (maybe more and more RM is adopted)

The study found an increase in the number of passengers bumped or denied
boarding because of oversold flights -- 1.01 denied boardings per 10,000
passengers last year, compared with 0.89 per 10,000 in 2005. JetBlue had the
lowest rate of bumped passengers; Atlantic Southeast the highest.

What should be my investment strategy?

2007-04-01T14:24:00.000-04:00

Recently, due to my wife work, I get a chance to learn those retirement plans. So how should I manage the investment? My fundamental idea is to combine investment and your future life plan. There might be two big decisions. One is when to buy a house and the other is when to go back to China. The two financial risks associated are housing price and exchange rate. I am not expert in either area. But we still can do something to decrease the risk without too much effort.Generally you want to buy a house at low price and come back to China when dollar is strong. But what if the opposite things happen at the time when your decision is made. So at the current time, I put a certain percentage of money into real estate and international stock market. If hourse price goes high and dollar become weaker when I purchase a house and go back to China, it might be offset by the increase in those two funds. Although I cannot perfect hedge that since I cannot put all the investment into those two sources ( currently since I expect subprime loan would drive the housing price down, I put much less money into real estate fund ), it is still worth to do.

An inconvenient truth

2007-03-31T21:38:00.000-04:00

Tonight I watch 'An inconvenient truth' again. It presents such a shaking fact - global warming. I suggest that everybody should watch this movie.

I should remind to turn down light every minutes.
Buy hybrid car for next new car.
Let everybody know TSP problem.

But when people tell you that our earth is going to destroy. What would you do? Take action to save it by controlling your desire or Do what ever you want to do before it is too late?

Trailer
http://video.google.com/videoplay?docid=2078944470709189270
http://www.climatecrisis.net/trailer/

Official website
http://www.climatecrisis.net/

Download
http://www.climatecrisis.net/downloads/

10 things should do
http://www.climatecrisis.net/pdf/10things.pdf

Chat Center

2007-03-25T19:34:00.000-04:00

There has been enomous study on the call center. MS has special issue on the call center back to 2004. Last week, I found there is potential another variation of call center. I call it chat center. There is a problem after I reinstall my computer. I find email response is too slow. Therefore I choose to chat to solve my problem. I find it is effective way to solve complex problem rather than email and phone. After my case got solved, I begin to ask that representives some questions and get some answer:

Question: How many customer do you serve every day. just curious
Answer: It depends, but as we try to sort and help out each of our
customers, it is normally 12 to 15 customers on an average.
Question: Is it possible that you can serve two or more people at the same
time
Answer: Yes, at times when there are chats waiting we have to attend two
customers at the same time.
Question: As I remember there is checkbox at the beginning to let me import
what kind of questions do I have. Does that mean you can solve all the cases?
Answer: We have all the tools that are needed to resolve all the hardware
issue, every customers issue is different and sometimes almost same.
Question: Just another question, how do you balance between two or more
people at the same time?
Answer: We attend at the maximum 2 customers and that too only if the
volume of chats is too high.
Question: What if both people doing something else so that no one respond
you in time?
Answer: This gives me more time to search and try to find out more
solutions.

Usually each server in the call center can only serve one people. But in the chat center, each server can serve more than one people. But now we cannot assume the service time is exponential. Otherwise your previous service time is useless. It is more like CPU case. But CPU job is deterministic. The order of serve in the CPU may not affect system performance that much. But human being is a totally different story. Which model is good, how to get system performance, how should representives balance service to more than 2 people?

Another application of chat center is MSN messenger. Some companies use msn to serve their customer. This might bring another whole bunch of interesting issues. How to balace of messenger service and current work? When to use messenger, when to use phone?....

We should always keep our eyes open to our daily life.

Where is Killer Application of OM

2007-03-25T18:30:00.000-04:00

Yesterday, I kept talking with different people. One is a graduated phd student from our department. He was used to work on inventory management in one Fortune 500 company. But the salary cannot compete with phd student from other field such as EE. We both feel that OM such as inventory theory (I doesn't mean those technology such as decreasing leadtime, since it has nothing to do with model) doesn't bring much improvement for the companies. For example, in the EOQ model, the model is quite insensitive to order quantity. And a lot of computation studies exclude cost such as purchase cost in order to emphasize the benefit of the model. Not to mention the restriction of the model. That might be low recognition of our field. I bet the number of people knowing IE would be less than 1/n knowing EE, ME... Why? Because those fields bring the more value to the society. People's live get changed by those fields.

Just like the slow progress of 3G, the lack of killer application obstructs development of 3G technology. The lack of killer application of OM/OR prevents smart people entering into this area and limits the amount of funding. We need more applications using OR/OM techniques to create great value to companies or even bring new companies.

Hope one day papers in our field are not just to improve the existence practice but also bring new model which companies would look into. Like CPLEX would send people to academic conferences. If more and more companies send their employees to OM ones. Also we have more master degree program other than MBA to train manager to have more quantitive competence. Then I would think our field gets recognized.

My research life approachs complete one more step

2007-03-24T11:33:00.000-04:00

From winter break, I am doing on something which I feel very good. And I spend a lot of time to do literature survey to make sure no one did similiar thing before. Then I began to do proof and write a journal-like notes with 20+ pages. But yesterday, I read a paper published in top journal which is a special case as my idea. Although I did something else and proof is different, I still feel very very frustrated. I received Larry's comfort email. After one night sleep, I feel a little bit. And I also chat with my college friend who is top researcher in my mind. It makes feel better also. Our chat history is attached in the comments. Chinese is mine and English is my friend.

The power of the blog

2007-03-21T18:17:00.000-04:00

As blog gets more and more popular, it becomes a source for the companies to advacate their products and improve their service. I remember there is a market research company which analyzes the blog posts and sell their analysis reports to its clients. Not it seems the companies take more preactive role here.

As I post the comments about dell product days ago. I receive the following email.

I am a customer advocate here at Dell headquarters in Texas. I found
your blogspot post about the problem you had with the motherboard on your Dell
Inspiron 600m (http://bsaint.blogspot.com/2007/03/balance-of-product-quality-and-warranty.html).
I wanted to check with you to see whether the replacement fixed the
problem. If you had any questions for me I would be more than happy to
answer them.

Today, the new motherboard is installed. It seems the problem is solved.

Responsible Authorship and Peer Review

2007-03-18T18:35:00.000-04:00

James R. Wilson provides nice introduction about authorship and peer review.
When you write a paper, we should remember

If you’re doing an experiment, you should report everything that youthink might
make it invalid—not only what you think is right aboutit: other causes that
could possibly explain your results; and thingsyou thought of that you’ve
eliminated by some other experiment, andhow they worked—to make sure the other
fellow can tell they havebeen eliminated.

He also gives several things to check as a reviewer. Although I never review any paper, it is good to keep those questions in your mind when you write a paper.

Are the problems discussed in the paper of substantial interest? Wouldsolutions
of these problems materially advance knowledge of theory,methods, or
applications?
Does the author either solve these problems or else make a contributiontoward a solution that improves substantially upon previous work?
Are the methods of solution new? Can the proposed solution methods beused to solve other problems of interest?
Does the exposition of the paper help to clarify our understanding of thisarea of research or application? Does the paper hold our interest andmake us want to give the paper the careful reading that we give toimportant papers in our area of specialization?
Are the topic and nature of this paper appropriate for this journal? Arethe abstract and introduction accessible to a general reader of thisjournal? Is the rest of the paper accessible to a readily identified group ofreaders of this journal?
Are the clarity and readability of the manuscript acceptable? Is thewriting grammatically correct?
Does the manuscript contain an adequate set of references? Is adequatecredit given to prior work in the field upon which the present paper isbuilt?
Is the material appropriately organized into an effective mix of text,figures and tables? Are data given in tables better presented in figures orin the text?
Is the work technically correct? Are the main conclusions justified by theexperimental data and by logically valid arguments? Are the theoremsstated and proved correctly given the assumptions? In practicalapplications of the theoretical results, do the authors check the validity ofthe underlying assumptions?
Are there gaps in the discussion of the experimental methods or results?If there are such gaps, can the closing of these gaps be considered (i)essential, (ii) desirable, or (iii) interesting? Are the experimental methodsdescribed in sufficient detail so that other investigators can reproduce theexperiments?
Have the authors explicitly addressed the limitations of their study—thatis, have they adhered to Feynman’s ideal of “utter honesty” and “leaningover backwards” in reporting their results?

One more step closer to finish all the courses

2007-03-18T12:25:00.000-04:00

I just submit my mid term take home exam for IP class. Now there are 2 final, 2 projects remaining before I finish all my desired courses to take. There are still some courses attracts to me. But I will only audit them and focus on doing research and make some progress.

A useful integral identity

2007-03-15T22:40:00.000-04:00

$(\int_{t}^{T}\sigma(t,u)du)^2-\int_{t}^{T}\sigma (t,u)\Sigma (t,u)du\equiv 0$ where $\Sigma (t,T)=\int_{t}^{T}\sigma(t,u)du$
This is the identity I found when I derive bond price under Q measure in the homework.

Lattice Search

2007-03-15T21:55:00.000-04:00

Suppose f(x,y) is a convex function. Now we want to find optimal integer x* and y* to minimize f(x,y) over integer set.
In continuou case, if we know x*(y), then we can just minimize f(x*(y),y) by taking derivative.
In discrete case, if we know x*(y), then can we say that -f(x*(y),y) is unimodular?

The balance of product quality and warranty

2007-03-11T18:50:00.000-04:00

Today Dell agreed to change the motherboard for my laptop after I complained that my laptop sometime automatically turn off when I use battary only. I also change the touch pad last year due its insensitivity.

Almost 2 years ago, I buy this 600m with half price. In order to do that, I need to find a configuration above 1500. Maybe unlike other people, I only choose to increase meory then I select 4 year at home warranty, since I cannot trust dell's quality 100%. And I don't like play game and don't do computation in my laptop. I think this laptop can still be used until my graduation if there is no big problem. Now this decision is rewarded.

But why dell cannot improve its production quality when he knows this product has 4 year warranty. I remember it cost about 250$ to increase from 1 year to 4 year warranty. But the change of motherboard is not that cheap. Consider anther 2 years to go, dell may lost money in my laptop heavily.

Companies like dell which has the information of customer order before its production, should be able to determine the suitable quality for its product based on warranty duration.

Inside the box

2007-03-10T12:54:00.000-05:00

'In my career, I've found that "thinking outside the box" works better if I know what's "inside the box". In music (as in life), we need to understand our pertinent history, and moving on is so much easier once we know where we've been' ---Dave Grusin (Award winning composer and jazz musician)

This is what I got from Starbucks cup yesterday.

XP to inventory problem

2007-03-07T22:45:00.000-05:00

Here XP is the product of microsoft. X can be substituted by I, L, NL, D, S....
Yesterday, I try to read the paper about multilocation and multiperiod model by Karmarkar (1981). Originally when I saw programming form. I think I can handle it without too much difficulty considering quite amount of time I devote to those XP classes. But it turns out that I am still lost finally. That's why everyone need to focus.

But I think using XP especially constructing dual problem will help us increasing the understanding of the inventory problem more. In his paper, he shows that base stock is optimal policy under certain condition for example transshipment. And it follows monoticity properities.

Although I think there is a potential mismatch of his SNLP form and the transshipment application, I really like his method. At least it can be alternative method to show that base stock is optimal under single location without using DP and those derivative stuff. I think that we need to see more alternative methods to prove the same important stuff. Wish one day I can see a paper with name 'simple/alternative proof of xxx in inventory theory'.

If someday, there are erratum for those major journals. Although there is notes but that is not enough. Even the smatest and most rigorous people will make errors all the time. Also it would be very very nice to someone to discuss with you about the same paper that interest both. The understanding would be deeper. You will truely understand something when you teach/discuss with somebody. Also the new idea is easier to come out.

Why lost sales is still unsolved problem

2007-03-07T22:32:00.000-05:00

These days, I read some papers about transshipment. There are two papers mentions why lost sales is hard to handle in general. This recall Zipkin's presentation about greatest unsolved problem in inventory theory. The reason is quite simple. If c(x) is convex, c(a(y)) is still a convex function when a(y) is affine function. But c([a(y)]^+) is not convex function in general. But it is still unimodal. After a little bit more thinking, another reason the authors not mention is that the sum of two unimodal may not be unimodal. That means unimodal doesn't perserve under expectation. So base stock policy may not be optimal policy. Hope I can a paper about lost sales to confirm this is the major reason to impede the discover of optimal policy form for lost sales.

OM in Health care

2007-03-07T22:14:00.000-05:00

As my wife works at consulting firm for a pharmaceutical companies, I begin to know a little bit about the health care system in US now. It is quite surprising to me when I heard that american people spend every 1 of 5$ of their income in health care from the radio, although I know this industry is huge. Then the next question is natural based on my background. What our OM/OR people can do for health care industry. Obviously we cannot develop the new medicine to cure the cancer. But there is some ways that we can improve the efficiency, which is all what we can do.

The first example comes into my mind is that an example about the quantity of cancer treatment in an LP book. The second example is how to minimize the total walking distance in the hosptial. But those are all about optimization model. We definitely can do more. The impact might be even more important than other industries such as airline which is consider the hot research area but suffer the hard time now.

Experience in Starbucks

2007-03-07T21:39:00.000-05:00

I have been in Starbucks frequently especially this week as a spring break. My working efficiency improves a lot since you can do nothing without internet access except the papers, books and winedit. I am quite satisfied with the environment there. You can stay all day long even if you just refill water after buying 1.85$ green tee. And when you are thinking, you can walk around. The more you go to the shop, the more familiar face you will see. Although there is no conversation, you still feel warm.

But today, I really feel tired at 3pm after finishing IP homework. However, there is no place for me to have a nap. I spend almost 2 hours to read all articles in today's NY time. From business to sports, from politics to alternative energy. This might be the first time that I read the newspaper so intensive after I leave China. I still remember the time that I fall asleep with whole bunch of business newspapers and magzines. I hope one day my english reading speed can be half as my chinese reading speed. Then I can be informed again.

The problem of unit in the relationship between backorder and waiting time

2007-02-24T11:44:00.000-05:00

The formula for the relationship between backorder and waiting time is
V[B]=lambda E[BW] + lambda^2 V[BW]

I think there is exact same formula in queueing theory about the number in the queue and waiting time. But the puzzle to me is the match of unit. Suppose the unit of B, lambda, BW are item, item/day, day. V[B]=item^2, lambda E[BW] = item and lambda^2 V[BW]=item^2.

If it was me to derive this formula, I would throw it into garbage after this check. But sometimes, we cannot trust common sense.

The difference and similarity between Poisson and BM

2007-02-20T10:30:00.000-05:00

The more I learn about BM, the more I feel there are strong connection between Poisson and BM. Last week, I ask a question. If we discrete Poisson process to small time interval, it behaves exactly as random walk with probablity lamda*h up by one and 0 otherwise. And counterpart of BM in discrete world is also random work with probability 0.5 up by 1 and probability 0.5 down by 1. Then I ask a question, what if we change the measure just like that we can change the BM with drift to BM without drift. Answer is obvious no, there are several differences.

discrete BM, up and down probability is fixed. the only scale factor is step size. But the scale factor of poisson process is probability up and step size is fixed
when we change the measure, we require that two probability measuresare equivalent which means the null space is the same. The null space of discrete one step of randome walke is R/{-1,1}. But poisson is R/{0,1}.

But when I ask Professor what is discrete case of BM with drift. Is it with the symmetric step size but different probability or different step size but the same probability or both or does not matter. I have not get answer yet.

IP homework is way too hard

2007-02-07T00:00:00.000-05:00

The second set of homework takes me two days with 16+ hours to finish and without leaving a logic hole.
The last two logic gaps I filled just now are
1. LP relaxation of TSP formulation with subtour elimination constrainted is strictly contained in TSP formulation with subtour elimination constrainted replaced by $u_i-u_j-nx_{ij} \leq n-1$. This invovle how to scale up the problem then scale down.
2. Find as many affine independent points as possible for $P={x \in R^+ | \sum^n_{i=1} x_{ij}=1 for i=1,...n and \sum^n_{j=1} x_{ij}=1 for j=1,...n}$. It would be easy to see the rules by beginning with n=3 and 4.

In homework 1 of IP class, the following question is challenging.
1. show that $max{x_1 − \sqrt(2) x_2 | 1 \leq x1 \leq \sqrt(2) x_2 and x_1, x_2 is integer}$ is feasible and bounded, but has no optimal solution. I use the contradiction to show that there is no optimal solution.

uniqueness of the fixed point

2007-01-29T00:09:00.000-05:00

In section 2.4.3 (Univalent mapping arguement) from 'game theory in SC analysis' by Cachon and Netessine, they claim that if best response function is one-to-one, then there is at most one fixed point(uniqueness of equilibrium). But why? It is possible that we can have x and y where x not equal to y such that x=f(x) and y=f(y). For example, f(x_1,x_2)=(x_2,x_1) is univalent mapping. But when x_1 = x_2=a in R, f(a,a)=(a,a). We can construct a trival game, player x and y write down a number, if x=y, then they both get 10, if not then their payoff is 10-(x-y)^2. In this game, the best response is f(x_1,x_2)=(x_2,x_1) .

Pros and cons of staying in a small university

2007-01-21T14:41:00.000-05:00

The most advantage is that the relationship between prof. and students are close. But sometimes you will be shocked. There is only one student registering real analysis class, so that the class is cancelled. It is hard to imagine that this could happen to the basic graduate math courses in those bigger universities.

Very limited courses are offered in Lehigh. After my three years in Lehigh, there is no open course about heuristics, the stochastic process class only using introductory level of book 'Introduction to Probability Models' by Ross. It takes me waiting for 2 years to have IP open again. And there are many other regrets about courses in Lehigh.

More guns can be added into my powder magazine

2007-01-15T14:11:00.000-05:00

Today, my wife gets and accepts her first offer. Now I can continue to my hobby to buy books. The books I plan to buy recently will be
Stochastic Calculus and Financial Applications
Nonlinear Programming: Theory and Algorithms
Theory and Practice of Revenue Management
An Annotated Timeline of Operations Research: An Informal History

Another new book in two days

2007-01-14T20:05:00.000-05:00

Today my wife's sister comes to my home with book 'Real and Complex Analysis' I ordered from dangdang.com. It only cost me 39RMB (5 dollar).

Why Toy's R doesn't have sample for people to try

2007-01-13T23:56:00.000-05:00

I also drop at Toy's R today due 25% gift certificate again. I try to get some gifts for my relatives and friends children. But there is no sample for most items for people to try except PS2 and Xbox 360. This setting loses the advantage to have a real store rather than online store only. Finally I get two idog mini which I already know what it is. But if I can try some other stuffs, I may buy more stuff. It seems that they don't know how to model customer behaviors

A new book comes to my book shelf

2007-01-13T23:50:00.000-05:00

Today, I use 25% gift certificate of Border to get 'Game theory' by Fudenberg and Tirole. It is must have book for people doing game theory in SCM.

Crazy Tuesday and Thursday Next Semester

2007-01-11T09:36:00.000-05:00

Finally I get my schedule for next semester. All the courses and meeting are concentrated in Tuesday and Thursday.
10:45-12:00 Advanced Stochastic Process II
12:10-1:00 IP Seminar (Th)
1:10:2:25 Real Analysis II
2:35-3:50 Financial Calculus II
4:00-5:00 Weekly Meeting with Larry(T)
7:00-8:15 Integer Programming
After crazy day, I will play tennis on Tuesday and Ping Pong on Thursday.

Wiki for Econ. Job Market

2007-01-06T12:16:00.000-05:00

Is anything similar in IE or B-school.

Assignment of interview time slot

2007-01-05T21:32:00.000-05:00

One of my friends from Econ. will graduate this year. He gets about 30 interviews during a conference at Chicago. Each interview lasts about half an hour. It is interesting to see how interviewees and interviewers agree on the time slot especially when there is no centralized system.

To queue or not to queue

2007-01-04T20:12:00.000-05:00

It is not about the book To queue or not to queue. It is about my experience today.

Today, I sit in the Wendy's from 10:30am to 2:50pm to read and think. When I realized that I soon need to have lunch, there is waiting line. Then it becomes an optimization problem. My hungry degree increases by time. I don't want to waiting in the line waste my time not doing meaningful things. The length of waiting line is changing. What should I do?!

It turns out that I was poor to make that decision. I waited until that I cannot stand more and joined a long line. What is the form of best strategy under this situation? What if I have perfect information or not? Is any similiar situation in the real business situation?

Martingale and NP Complete problem

2007-01-02T13:47:00.000-05:00

Last semester, I have two courses related to martingale. But how can it be used in inventory related problems. So far my limited literature reading hasn't touched any paper using martingale properties. I guess the reason is that it is not easy to find a martingale in inventory problem.

Today when I read 'The logic of Logistics', I found an application to use martingale difference sequence to develop average case analysis for Bin pack problem. Although it is not close related to inventory problems I am interests. It is a good start.

The reference is
Martingale inequalities and NP-complete problems

Wrong impression: any NP complete problem has no finite worst case bound for any heuristic

2007-01-02T13:28:00.000-05:00

In network and graph class, we learn that there is no finite worst case bound for TSP. But I thought there is no finite worst case bound for any NP complete problem. Because of this wrong understanding, I argued with a speaker during Informs about his speech about bound of one multi-echolen problem. Of course, he cannot agree on my wrong opinion.

I realize this understand is wrong until today when I happen to read the bounds on Bin-Packing problem.

live as a pig at the end of 2006

2006-12-31T23:05:00.000-05:00

Maybe florida is too warm. I caught the cold right back to Bethlehem. Now everyday, I sleep, eat, sleep. Today, I finally try to do something different and played monopoly. As usually, I defeat computer players with my perfect preformance in stock market. If it were real money, oh...

Hope tomorrow 2007, I can recover and go back to normal status.

Two OR applications in Florida Trip

2006-12-28T09:35:00.000-05:00

The first one is queueing theory.
When I was in Magic Kingdom, I found an interesting application of queueing theor - Fast Pass. The procedure is the following. 1) When you see a waiting line is quite long at one attraction say Space Mountain, you can insert your ticket in one machine to get a fast pass which indicates you return during the specific time usually one or two hours laters. 2) Then when you return after playing other stuffs, you can go through the fast pass line.
There are several key points. 1) Once you get your fast pass for one attraction, you cannot get fast pass for another attraction until your current fast pass expire. For example, if I get fast pass for Space Mountain from 1pm to 2pm. I can only get fast pass for Splash Mountain after 2pm. 2) The number of fast pass in one attraction must be limited. Since the available time for the fast pass is different at different attraction, the hotter the attraction is, the longer time difference between available time of fast pass and current will be. In the last attraction-Jungle Cruise we go at 5pm, its fast pass is unavailable. We need to wait 40mins to play.
The decision to make for the visitor. 1) What is the sequence of attraction I should follow. 2) Which attraction should I get the fast pass.
The decision to make for the Theme Park. 1) How much fast passes to available at each attraction and each time periods.
Game theory might not be suitable in this case. Probably, something like priority queue plus abandon and other customer behavior can be used to formulate the problem.

My experience of fast pass in Unversal Studio at CA is different. We got the fast pass after we experienced a half hour technique difficulty at Jurassic Park. Then we can get into fast pass line for every attraction. But actually, you need to spend 30$ to get one fast pass. So the decision for the theme park is how much fast pass to offer and how much to charge.

The second one is network flow. It is very wise decision to buy a GPS just before this trip. We don't need to print every map. And we can avoid the traffic jam in the high way by going through the local. It was very crucial when we tried to catch up our flight back. And the calculation of route is very fast, it is much faster than O(A) I learned last semester. But actually it can only calculate the route from two points. If I want to go several points a day, I need to manually choose the sequence of points I need to goal. If GPS can provide the function of solve small TSP problem such as 5 points, it will be wonderful. Since real map satisfy triangle inequality, there is heuristic to get solution within 1.5 time of optimal solution. So the calculation time won't increase a lot.

Final week is over

2006-12-16T00:13:00.000-05:00

This should be my second last final week. Now I am enjoying the final quarter of game between Rocket and Laker, two of my favorite NBA teams. As usual, I don't have mood to study once there is two or three weeks to final week. This year this mood is especially strong cause Rocket comes strong this season. After I finish the trip to Florida, I should use my time sheet again to do my time management which I haven't done after INFORMS.

Textbook: Convex Analysis and Optimization

2006-12-14T12:08:00.000-05:00

Today I want to search how to prove a function is unimodal. I get the following book, Convex Analysis and Optimization by Dimitri P. Bertsekas. What I like Dimitri P. Bertsekas most is that he always post his class slide accompanying his book. Another example is his book Dynamic Programming and Optimal Control.

How to Bring Our Schools Out of the 20th Century

2006-12-14T10:17:00.000-05:00

Time shows the views how American people would like education to be in the 21th century. Let's compare to Chinese education I received

Knowing more about the world. (China: Almost all high school will teach English and we learn history of world. USA: where fewer than half of high school students are enrolled in a foreign-language class and where the social-studies curriculum tends to fixate on U.S. history.)
Thinking outside the box. (This is weakness of Chinese education. But it seems with no child left behind(NCLB), American goes to Asian type education more and more.)
Becoming smarter about new sources of information. (This one is hard to tell. Only thing I can say is that more knowledge I have, the smarter I can deal with new information)
Developing good people skills. (This is also a weakness of Chinese people. There is a saying that 'three monk don't have water'. And I found the counterpart in English:Everybody's business is nobody's business. So it is also a problem in USA. I think it is due to how people get measured. Maybe in USA, there is fairer measure of team work than China. For example, almost every teacher ask students to put down name of people who help their homework)

One surprising thing is to see the following question for a second grade student: How many ways can you combine nickels, dimes and pennies to get 20¢? This question can be solved by recursion. It would be interesting to see how teachers tell students the way to solve the problem.

Three Major Assumptions in Marketing and my customerized assumptions for MS/OR

2006-12-12T15:27:00.000-05:00

It is interesting to know the major assumptions in marketing.

The customers are complete. It means that they can compare any two items. 'complete' here is similar to the definition in the math. There is no hole
The customers are transitive. If the customer prefer A to B and B to C, then they prefer A to C.
The more choices, the better.

Is any major assumption in MS/OR field? I never see anything like those three major assumptions in Marketing. But we can have similar one.

The decision makers are complete. It means that they can compare any two solutions.
The decision makers are transitive. If the decision makers prefer A to B and B to C, then they prefer A to C.
The larger decision variable sets, the better result the decision makers can get.

Don't need to get up at 8:30 every weekday now

2006-12-11T17:02:00.000-05:00

This week is the final week. During the whole semester, I have classes beginning around 9:00 everyday. Now everything is over and I can get up whenever I am awake.

Bless T-Mac!

2006-12-09T22:29:00.000-05:00

Although YM got 23pt in the fourth quater and hit the clutch shot, it cannot make me happy. It is because T-Mac got hurt his back in today's game with Wizards. Rocket has very good beginning of this season. All we need is to wait for T-Mac get recovered with 80% degree compared with his status vs Dallas during the playoff at 04-05.

I really like T-Mac this season. He changes his style from scoring to controling the game. His basket IQ is a key for the success of Rocket. With T-Mac and YM, I can spend almost 3 hours every game day. But now it seems Rocket will go to lottery again with this injury, even though YM has very strong performance this season. Let's hope it is just minor injury and nothing is related to old one.

My current efficiency drops so fast!

2006-12-04T00:04:00.000-05:00

Maybe towards the end of the semester, I become stupid. Or I look too much NBA and related news. My instinct is gone very quickly. Even one set network homework takes me almost the whole sunday. And I still haven't finished it. I am looking foward winter break now and need to control online time.

Future is Present

2006-12-01T23:33:00.000-05:00

This is one of the beautiful thing you can learn from math course. In financial calculus class, the first half class is devoted to binomial tree model. In that model, you rebalance your portfolio in the near future (Delta t_i) based on the information you know during Delta t. But in continuous model, Delta t becomes delta, future collapses to present. You need to do rebalance all the time. If not, you will be arbitraged by others.

If you don't want others to arbitrage you, you'd better to know future is present and try you best continuously.

tricky tricky airline price

2006-11-29T11:00:00.000-05:00

I know a little bit about revenue management, such as littlewood's rule. In my mind, the price of air ticket should stochastically increase. When I try to book the tickets to Orlando from the southwest during the winter break, the price change of air ticket is quite dymatically as I can imagine. The lowest one way price in my case is $59. This price will be randomly available at each day. Moreover, the number of available seats also change at each day. That might be due to change of reservation by the other people. But it may not be entirely true.

Yesterday when other couple reserved one ticket at one time slot(depart at 12:00) and tried to reserved anther one at the same time slot, it failed since there was only one slot remaining before they reserved. So they changed to 10:00 now. Today that slot is available too. It seems that someone change or cancell their reservation yesterday. But it is not just that. I changed mine 12:55 to 10:00 in order we can go togather. But I cannot find available seats for 12:55 today(Yesterday 12:55 is unavailable). One possible explanation is that demand at 12:55 is stochastically higher than demand at 10:00 with respect to the same price. So southwest move the tickets available from me to higher price. But I don't think this is the whole story. What is the control mechnism of air ticket price at southwest is still a puzzle to me.

Btw: southwest allows you to change your reservation without any penalty. That is why "Last August, low-cost king Southwest Airlines carried more passengers than any other U.S. airline, the first time an LCC has claimed the top spot." reported by Time

Stochastic processes@wikipedia

2006-11-27T10:57:00.000-05:00

I haven't seen most of them. A long way to go!

Poster presentation next year at INFORMS charpter@Lehigh?

2006-11-17T17:04:00.000-05:00

Today Lehigh's INFORMS chapter had its first meeting this semester. Basically we discuss who will be invited as INFORMS chapter seminar speaker next semester. After that, we need to figure out a way to spend increased budget. There is one idea comes out during the discussion. That is we will provide T-shirt having ISE logo. I like this idea a lot. At least we need something to connecting people togather.

Right after meeting, I get an idea about create poster presentation next year. Now in our department, those optimization people have very strong connection. They have their own CORAL and IP seminar. But the connection outside optimization is quite loose. So hope this poster presentation will let people know what the other people do. And every poster presenter should wear the ISE T-shirt.

I also get idea to create subdivision leader in our chapter, say, optimization, stochastic, finance, scm&rm. Each leader should arrange 3 weeks long seminar. This is just coming from IP seminar yesterday.

Maybe we can also create the connection between IE and other department student, so that we can see what cool stuff we can do for other fields.

And just now, I send an email to our president about giving a talk to the first year phd student of the career path of phd. It is very often to mention in the company but seldom done by university. I guess after that we can also create Q&A page for all IE student.

I believe by doing more stuff like that. We can create the best INFORMS charpter ever and sell OR to more people.

If we can get enough people joining us, we can also make the newsletter once a semester. Maybe some short introduction about what happens to every member in our chapter. What had done last semester and what we will do next sememster.

Keeping your direction, don't just follow fashion

2006-11-11T17:49:00.002-05:00

This Friday, Kevin Shang gave us a talk about multi-echolen inventory problem. It is very classical problem but there aren't many people focusing on this area. And the barrier to this area is higher than most other areas related to inventory theory. So most people like to chase research 'fashion'. Somebody told me that a few years ago, sessions of SCM are full of people. But in this Informs meetings, RM seems more popular. A lot of people talk about dynamic pricing etc.

But Shang sticks to multi-echolen inventory problem after he passed his qualify exam. And he told me that he will continue to explore this area. It should be appreciate. Although our field just like a branch of applied math, it doesn't mean that everybody should just apply similar math technics to different applications. Once the math is too hard to go on in one area, then everybody shift their research interests. I don't mean that we shouldn't do so. In contrast, we need some people to do so to spread out our field. But what I mean is that we need some people to develop some unique math methods to solve those unsolved problems. Like developing in physics can push math forward. Why won't our IE people do the similar things? (Btw: I am not talking about those optimization people in IE fields)

Actually, I am a guy, who cannot focus on one thing for a long time. I like to find tons of applications, which can be easy in math. But I will find an area which can attract me for at least ten years. Of course, I will do other things on the same time.

What should I do in the next Informs meeting?

2006-11-10T20:47:00.000-05:00

Since I already get to know some friends from this conference, I hope I can meet each one individually and have deep conversation. Also because, I will have more time to do research from the fourth year, I hope to seek more opportunities to work on the different subjects with different people with their own expertise
Prepare some deeper work to present. That's the only way to impress people, so that they are willing to work with you if possible.
Get to know job market
Create the chance to talk with those faculties
Last but the most important, live at a hotel nearby the convention center. I run into trafic every day and get too less sleep every day during conference. And I haven't recovered yet.

internal force(nei gong) or external technic(zhao shu)?

2006-11-10T19:42:00.000-05:00

I went to an Informs conference from Nov. 4 to Nov. 8. Everything goes well. Meeting and making friends is the most important thing for me since our department only has very few people sharing the same interests with me. I need to extend my network. Second thing is to present my work. It is ok since nobody left during my talk and I got two questions and handled well in my opinion.

But there is one fundamental question occuring several times during the conference. On one hand, we need to have good papers before we graduate. It means that I need to work hard on generating papers, which implies that I need to read a lot of literature. As Long told me that we need to accumulate several tricks which are used by different authors again and again in their papers but no one collect and publish those as a reference book. This is absolutely right. Even those great mathematicians has only a few tricks. On the other hand, I like to take as much courses as I can since I believe only when you have complete skill set, you can consider problems from different angles which gives you more insight. Also you can do something crossing the different fields.

In Chinese Wuxia novel, there are two kind of persons. One is working on internal force(nei gong) . It takes a long time to master. The other person is working on external technic(zhao shu). It is fast to learn but easy to hit the bottleneck. In the short run, the people working on zhao shu can easily beat the previous type of people since the stuff they learn is more practical. But in the long run, the previous people can outperform the people only having zhao shu since it has more power.

In this conference, when people start to talk about those 'famous' people's name, I feel a little bit embarrassing since I cannot recognize a lot of those 'big' names. I start to think whether I spend too much time in taking course other than reading related literature. For example, the next semester, I plan to take two courses and audit another two courses again. But I stick to finish those courses since I won't get any time to do so after I graduate. And only when you have enough nei gong, you can pick up those tricks quickly. I am not worried about my ability to come up ideas. Actually, during this Informs, I generate several new ideas. For now, I need to focus on developing those nei gong and try to work on only one or two ideas. After next semester, I would devote most of time into research but not now.

MSOM 2006 submission statistics

2006-10-29T14:27:00.000-05:00

Till the latest data, there are 154 submission for the regular issues and 36 submission for the special issue.

For the regular issues, there only 4 papers(2.6%) get accepted. Among these 4, there are 2 papers get accepted without going through review process. 8 papers(5.19%) get minor revision and 35 papers(22.73%) get major revision.

For the special issue, 4 papers(11%) get major revision and 4 papers have got the decision. All other paper got rejected.

You can access data from http://msom.pubs.informs.org/msomdata.xls

The accpetion ratio is very low. MSOM currently publishs 4 issues a year. And each issue contain around 6 papers. So the acception ratio is around 24/(154+36)=12%. That means around 2/3 major revision paper will get rejected finally.

Lehigh ranked 18th undergraduate B-school

2006-10-23T14:14:00.000-04:00

It is a little bit surprising to see Lehigh can get 18th rank in that category. ANyway, it is still a good news.

Beamer's columns

2006-10-21T09:30:00.000-04:00

\frame
{
\frametitle{Interview}

\begin{columns}
\begin{column}{0.5\textwidth}
\begin{itemize}
\item current symptoms
\item past medical history
\item family history
\item social history
\item review of systems
\end{itemize}
\end{column}

\begin{column}{0.5\textwidth}
\begin{itemize}
\item detect hypertension
\item recognize emergency
\item recognize reversible causes
\item find chronic target organ damage
\item discover co-morbid conditions
\item plan therapy
\end{itemize}
\end{column}
\end{columns}
}

You can have as many columns, of whatever widths, as you like. Note that
within a column, \textwidth will correspond to the column width, not the
width of the beamer slide, so it works pretty much like minipage.
Material within columns is vertically centred, and you can have part of
the slide in normal (single column) format, and other parts in more than
one column.

Insert List into Tabular in Latex

2006-10-20T21:30:00.000-04:00

Here is the code. But I still cannot figure out how to do under Beamer. The ultimate goal is to insert the Figure into the left top corner and put the list on the right.

\documentclass[12pt]{article}
\makeatletter % Yes, it's horrible
\def\spacehack{%
\@minipagetrue% % Disables spacing above
\expandafter\everypar\expandafter{% % Add stuff to \everypar
\the\everypar% % Do what was there before
\@minipagefalse% % Clear the mystic flag
\everypar={}% % And reset \everypar
}%
\let\@oldbs=\\% % Remember old \\ command
\def\\{\nointerlineskip\@oldbs}% % Turn off final odd spacing
}
\makeatother

%
\begin{document}
%
\begin{tabular}{p{40mm}p{40mm}p{40mm}}\hline
& Advantages & Disadvantages \\ \hline
Method 1 & \spacehack\begin{enumerate}
\setlength{\itemsep}{-\parsep}
\item item one
\item item two
\item item three
\end{enumerate} & \spacehack\begin{enumerate}
\item item one
\item item two
\end{enumerate} \\[-12pt] \hline
Method 2 & \begin{enumerate}
\item item one
\end{enumerate} & \begin{enumerate}
\item item one
\item item two
\end{enumerate} \\ \hline
Method 3 & \begin{enumerate}
\item item one
\end{enumerate} & \begin{enumerate}
\item item one
\end{enumerate} \\ \hline
\end{tabular}

\end{document}

How to be a good lecturer?

2006-10-19T17:20:00.000-04:00

You need

Have interesting problem (not to be too hard and too specific, especially when your audience background is mixed
Always talking something that is easy to understand.
More figures than words and formula
I will add more here

How to be a good listener?

2006-10-19T17:15:00.000-04:00

From those experience, I see there are several keys to be a good listener.

Try to understand the main point
Try to answers all the questions other people ask
Try to think all possible ways to attack their theory
A good lecturer is neccessary

How I remember those speakers' presentation(2004, Fall)?

2006-10-19T15:18:00.000-04:00

Yesterday I read through the advice to the young scientist. It gives some advice about how to do a good talk. The first advice is 'Every lecture should make only one main point. ' And the fouth advice is 'Give them something to take home.'. What can make people remember you. Here is the answer.

It is not easy to follow this advice. It is easier to state what features of a
lecture the audience will always remember, and the answer is not pretty.
I
often meet, in airports, in the street, and occasionally in embarrassing
situations, MIT alumni who have taken one or more courses from me. Most of the
time they admit that they have forgotten the subject of the course and all the
mathematics I thought I had taught them. However, they will gladly recall some
joke, some anecdote, some quirk, some side remark, or some mistake I made.

So I want to recall all the seminar in our department I attend to see what I remember from those talks. Then I may know how to improve my presentation. And how to be a good audience. Here I need to clarify. The talks I don't remember now may be due to that I don't have enough knowledge at that time. I begin from 2004, Fall which is my first semester here.

2004 Fall, I attended all the seminars. The following is the talks that I still have an impression.

08/27/04 David Drake Lehigh Phd student: This is first seminar I attended. The only thing I still can remember is that he thought HongKang is nice place.

9/17/04 Suvrajeet Sen. Program director at NSF. This was the time that every professor was going to take notes. I remember that Dr. Ralphs and Linderoth asked some questions about software patent.

11/5/04 Aurelie Thiele. Lehigh. I know newsboby model can also extend to revenue management. But the terminology is different. Basically, for the simple concave problem, the optimal solution is achieved at marginal cost equal to marginal revenue. In the revenue management, if we have high fare class and low fare class. The number of seats reserved to high fare class is achieved at expected revenue brought by high fair class is equal to expected revenue brought by low fare class.

12/3/04 Krishnan Anand, Upenn, There was an interesting situation that sometimes information may cause increase of inventory. I also remember there is an question related to Wal-mart. Later on, I find a good explanation and email it to Prof. Anand.

An example of converge in probability but not almost sure converge

2006-10-17T18:38:00.000-04:00

This is one homework question in stochastic process class. I cannot come up even one example by myself. First I use goole, nothing valuable is found. Then I http://books.google.com/. I get a book named 'Counterexamples in Probability and Statistics'. The example is X_n=H_n, where H_n is Bernoulli variable with probability 1/n to be 1. X_n converge to 1 in probability. But Prob(H_n = 1 infinity often)=1 by Borel Cantelli Lemma.

2 hours to clearn up the inbox

2006-10-15T18:41:00.000-04:00

I store all the emails in the box and only delete some when I get the warning about storage. Last time I did clearning is last semester. Today I spent 2 hours to move every email to the proper local folders. It is a good chance to see what I have done since the end of last semester. Although the outcome of research isn't that breakthrough, I am still satisfied with how I deal with those multi-tasks. Now I have fewer things to do. I can put more emphasis on each thing I am doing.

Advice for the Young Scientist

2006-10-15T14:23:00.000-04:00

Another related article mentioned is Ten Lessons I Wish I Had Been Taught

My presentation@IP seminar: "rondom thoughts about optimization"

2006-10-13T12:31:00.000-04:00

Yesterday I did my first outside class presentation Since my research is not mainly about optimization, I have to make up some stuff to fit the audience's interest. Also I hope that my talk can create some coordination between students in our department. So I just provide several of my ideas which I think that might be attractive to other people.

The result of presentation is OK, at least people don't feel bored during that 50 min. But I am sure that I don't achieve my target fully, since no one shows strong interests. Larry told me that my confidence overcomes my spoken English. Prof. Huang shows a different approach to solve one of my problems about LP and regression after presentation, where there is no need to change matrix A but only constants on RHS. Also I find another way too. We can just change one column instead of replacing the rows. Prof. Linderoth also provides me how do sampling in multi-stage stochastic programming. Prof. Ralphs reminds me how to get dual solution easily. Ash shows interest on my network formulation and I explain to him about bootstrap idea further. Zumbul told me that she has her own problem and want to coordinate with me. Finally my wife told me that I put too much contact with those professors and lose contact to the students.

Finally, I should say I need focus. After that I hope get any coordination with other students.

I have three SCM books now

2006-10-10T21:02:00.000-04:00

After struggling for a while, I final decided to buy two more SCM books. They are

The logic of logistics
Foundations of stochastics inventory theory

Plus I already have

Foundations of Inventory Management

Also I copy

Inventory Control

What else do I need to collect?

The best deal for the Economist ever

2006-10-10T20:59:00.000-04:00

Today when I got books from Amazon. I found such good coupon. There is only $100 for 51 issues of Economist plus $25 Amazon coupon. I wish that I could have to read.

Prediction of Nobel Prize in Economics 2006

2006-10-06T10:34:00.000-04:00

I guess it would belong to Bhagwati, Dixit and Krugman since their topic is more interesting to me.

measurable or nonmeasurable

2006-10-05T22:54:00.000-04:00

Now I am doing Real analysis homework, it is hard to imagine those strange sets. The course till now is all about trying to show that close properties about measurable set and functions. It is hard to come up your own counter example and deal with those 'strange' stuffs like Cantor, Borel, nonmeasure sets.

Luckly by keeping doing the homework even though the score is as normal as my other courses, I am still not lost in the class. It should be considered as the hardest course among all ones I take and audit. Maybe it is time to go office hour next time.

Arby's coupon

2006-10-04T13:27:00.000-04:00

Yesterday night, I came to Arby's to have my dinner to save my time. I use the coupon Arby's sends out to me every month. I just raised my hypothesis before I order:
People eat insides are more price intensive so that more people use coupon than those who drive through.
Supposing my hypothesis is true, if Arby's send out the coupon to everyone no matter whether they prefers to eat inside or drive through, then it may lose potential revenue since it increases the chance of using coupon among the driving through persons. After I observe during my dinner, four transaction occurs all including coupon. That increases my confidence on my hypothesis.

A naive way is to replace current strategy with only putting coupon inside the store.

Again, we can build a model to help Arby's to improve their revenue

oonumerics.org

2006-10-02T19:41:00.000-04:00

It provide information about different package in the following catogories

Linear Algebra
Arrays and Images
Neural Networks, genetic algorithms, machine learning, data mining
High-Energy Physics and Quantum Chemistry
Multiprecision, arbitrary precision data types
Differential Equations
Automatic differentiation and interval arithmetic
Visualization
Graph Theory/Combinatorics
Language interoperability/scripting
Transforms
Optimization
Miscellaneous
Tools

But there are several packages in each catogory. If there is someone providing comparison, it will be great.

JabRef

2006-09-27T23:29:00.000-04:00

A quite nice .bib document organizer.
It makes me comfortable to write down my comments to every paper I read.

Power sum

2006-09-19T23:25:00.000-04:00

sum(i) = n(n+1)/2
sum(i^2) = n(n+1)(2n+1)/6
sum(i^3) = n^4/4 + n^3/2 + n^2/2
...

A site contains optimization online resource

2006-09-19T11:21:00.000-04:00

When I check the statistics for COR@L today
I get the following link, which provide some good online resource for optimization.
http://plato.asu.edu/sub/tutorials.html

The 3rd weekly report

2006-09-17T23:45:00.000-04:00

This week, I work for 56 hours. Due to easier homework of Graph and Network, I only spent 28.5 hours in study. Also I make a little progress on one of my research project.

Beside work, I watch the football game between Lehigh and Princeton. I sit too low so that sometimes I cannot clearly catch what was going on in the field. Next time I will sit high enough.

I also watch the film The Illusionist. It is a wonderful movie. Originally I think I would fall in sleep during the movie. But I cannot stop enjoying it and have no chance to have a nap.

(-1)^(1/3)=0.5000 + 0.8660i ?

2006-09-12T22:05:00.000-04:00

This time, I meet another problem about math operation in software.
In matlab, when you try
(-1)^(1/3)
What result do we get. It is not -1 but 0.5000 + 0.8660i
You need to use nthroot(-1,3) to get -1.

This time, excel gives me the answer I want.

I am so frustrated now. This time I use more than 4 hours to debug my program which contains similar operations in my matlab code.

Do I know too little about basic math or Matlab provides too much.

My 2nd weekly report

2006-09-10T23:59:00.000-04:00

This week, my course work increase 1 hour. It seems that 30 hours per week is average workload 4 courses taught all by professors from math department. They usually have more homework than IE courses. But it is worth doing that.

But research work drops to 23 hours per week. That's because I finish all the coding work for one of my project. I need to increase it to above 25 hours a week.

Also I have played tennis twice this week. I am still seeking the feeling about backhand, which I lost for a month after I changed my habit and cannot go back.

EXCEL vs GOOGLE

2006-09-10T23:41:00.000-04:00

When I try to verify the correctness of a sequence. I need to compute the following fomula

=-1*((0.4^4-3*(0.4^3)+0.4^2)*10+(0.4^3-2*(0.4^2))*15+(0.4^2-0.4)*20+0.4*25+30)+100*(0.4^4-2*(0.4^3)+1)+50*(-0.4^4+2*(0.4^3))

Excel gives 66.144. I takes me more an hour to debug. Finally I find it is not my fault

But google gives 63.58400

Actually google gives the correct result.

How can I trust excel anymore?

I still remember two year before Excel can generate negative number from [0,1] random variable. Luckly I get the patch to fix it. This time, I don't how I should do.

chalk and slide

2006-09-04T21:35:00.000-04:00

This semesters, all my attended courses are taught from Math department. Those professors like to present their material by chalk on the blackboard. In contrast, most of my IE courses are taught through slides. But I have to say that I love chalk style. You can see how the professor thinks from chalk but not from slides. Also when you take down the notes, you not only use ears to listen, but also use eyes to look and hand to write down. The more 'sensors' you use during the class, the more and the longer you can remember.

One possible reason for the IE professors to use slides more than chalks is because IE professors uses more computer than Math professors. So the technology not always create better way to learn. But slides is healthy way to teach since the more chalk you use, the more chalk ash you absorb. Also use slides may have huge scale of economic. If let me choose which way I use. I would be half and half. Just like last semester nonlinear programming class. Use the slides to show main ideas, use chalks to show the details.

My 1st weekly timesheet

2006-09-03T14:13:00.000-04:00

Although, it isn't 5:00pm at Sep. 3, I like to post my 1st weekly timesheet. The total hours on study are 29.5 and on research are 31. I am quite satisfied with my work in the 1st week. I will go on this self monitor mechanism

Andre Agassi & NP-complete

2006-09-01T00:49:00.000-04:00

I just finished watching a unbelievable tennis game. No match can be better than this one. All the unbelievalbe games have equivalent elements, like all NP-complete problems are equivalent.

Poincaré conjecture, Perelman, Fields Medal

2006-08-27T18:44:00.000-04:00

There are a lot of quarrel about how much credit everyone involving to solve Poincaré conjecture should get. But the most important person, Grigory Perelman, a Russian mathematician, deline which I think is more valuable than Nobel prize.

It shows that the fundemental research still need genius and have genius. Not all the work need to be done through cooperation and fund.

There is good post from MITBBS.com(A chinese forum for people in North American)

It's time to change my own time clock

2006-08-26T00:28:00.000-04:00

This new semester, I have the classes beginning around 9:00am every weekday from next week. I cannot believe the summer was over. The time is passing so quickly. Unfortunately, I cannot achieve the goal I set at the beginning of this semester except auditting the class and playing tennis. I really need to push myself hard this semester.

I take two courses: network and graph, financial calculus. Also I am going to audit: real analysis, advanced stochastic process.

But my eagerest goal is to finish two projects on hand so that I can put my time into real research. For my research, I wanna to do more theoritical side than computational side.

There is something I want to get rid of. I am going to shrink my online time. It really wastes my time. To accomplish this target, first not bring my laptop into school unnecessarily, second not open computer after I get up and before I go to bed, third do not go to code unless it is worth doing, fourth don't try to read basketball information everyday especially when new NBA season begins.

At the same time there is something I am going to get used to. Since I am moving to a two bedroom apartment recently, there is a study room for me. I can study home now. Also I get a new mattress, it is so comfort that I hope I can read a paper everyday on the bed before I fall into sleep.

Hopefully, I will increase my ability to manage my time and be more productive. I am during the process of designing a weekly agenda. It needs the balance between work and life. And I am going to fill out my timesheet everyweek as I need to do as an employee. Also I would send my weekly report one day before I meet Larry with detail checklist of what I am doing during the work. I learn it from Wasu. He told me that is really efficient way. I will post the formats of my timesheet and weekly report later. And publish the result of my timesheet and the number of jobs I do every week.

In the ideal situation, this gonna to set my high expectation more likely to achieve.

The connection of EE and SCM

2006-08-01T21:43:00.000-04:00

Finally, I found a connection between EE ( my undergraduate major ) and SCM ( my current research interest ).

Yesterday, I call my 4 year roommate at JiaoTong Univ. , who is now at EE@UCSD. I happen to ask whether OR can be applied into his area. Originally, I just want to advocate the usage of OR and provide him a potential tool. It seems no many people use LP, IP, NLP to his area. After I push him think hard, he gave me a very interesting problem in the physical network design. He thinks that using large scale optimization model to solve real time network flow is not his favorite. i.e. he don't like the idea that doing optimization at server level and then telling what the router should do. Instead, he like to solve the problem at the router level and the objective is not far away from the global control.

That's the exact idea of centralized vs decentralized, which is one of the key concept of SCM. But the research on the complex supply network is far from mature. So I cannot give him a plausible method during the phone call.

Also his current internship at Motorola is to implement at heuristics to solve IP at cell phone. Although he uses the different langurage, it seems branch and bound is way to complex for the chip of cell phone.

Anyway, it is always interesting to find the people at different area trying to solve the similar problem.

Personal cost minimization on air ticket purchase

2006-07-22T12:47:00.000-04:00

This week, Time privdes a very interesting website to help customer to make the decision about 'To Buy or not to buy' air ticket.
http://www.farecast.com

Etzioni's site Farecast.com
original name was, of course, Hamlet--provides the lowest fare on a route, a
90-day price history and, using a novel predictive algorithm, a tip to "buy now"
or "wait," along with a figure indicating how confident Farecast is in its
advice. (Flyers buy directly from the airlines.)

It is surprising that this kind of website is not developed by the OR people (especially revenue management people ) but a CS professor Oren Etzioni from University of Washington.

OR people need to establish much more sucessful story to advocate importantce of OR. Like a disucssion raised in the Coin-OR workshop, we need more credentials to demonstrate the impact of OR. If OR can become HOT in job market, the pressure of academic people can also release and they can focus on the question they are more interesting but time consuming.

DIMACS workshop on Coin-OR

2006-07-20T22:01:00.000-04:00

This week, I attend DIMACS workshop on Coin-OR. I am not pure optimization people and have not implemented any project based on any package of Coin-OR before. But I am really astonished to see those really really busy people devotes their time to the open source software on the optimization, even if it may be necessary for them to develop code for their own research. This fresh experience brings me to a new world. Besides this, this trip gives me a chance to know other people in our department better.

During the workshop, I get to know a person from Archstone-Smith working on the revenue management. He came here and try to find a package to replace Cplex which their software is based on. But it turns out that he is a little disappointed on it. The reasons are warmly discussed in the last presentations. They may be the common concern about open source software, legal and support issue. It would be nice to see a company making money by supporting Coin-OR. The problem is whether there is such demand? Then it leads to the issue about awareness of OR either in public or among other research areas. As Stephen Nash point, there are big and interesing questions in other science fields such as how universe begins. If everyone include me know the biggest problems in OR area, the future of Coin-OR would be bright. Otherwise, it is just a tool of researchers.

I hope there is one day a book introducing how to model OR by using Coin-OR through a well spread tool such as Excel like VBA for Modelers: Developing Decision Support Systems Using Microsoft® Excel. And the reader reviews is over 20. This is similar to Leo Lopes presents at the last part of his presentation to show how to use web service through VBA.

The presentation file can be found in workshop website.

The world cup champion predicted by UBS

2006-07-10T00:09:00.000-04:00

UBS reports shows Italy would win the champoin. Maybe gamesters should not just read the sports news but also business new.

Many influencing factors
Along with "football fever", the team around
UBS Wealth Management Chief Economist Klaus Wellershoff was interested to find
out what variables are important in predicting World Cup success. In doing so,
they discovered that many things that appear to be obvious are, in fact, not
crucial to winning the World Cup. An example of this would be the size of a
country's population, which is often incorrectly correlated to the amount of
potential athletic talent. The FIFA rankings, which list the top national soccer
teams, also prove to be of limited use when it comes to making a prediction: the
FIFA list compiles the sporting success of the individual teams but assigns
equal value to all wins, no matter how strong the opponent.

GLPK under windows

2006-07-07T11:07:00.000-04:00

I experience a hard time to use GLPK under windows. 2 months ago, I can run the program under vc6. But yesterday, linkage errors persist to appear. Finally, I get the answer that I didn't specify glpk.lib in the project settings. But last time, I also not set it. Anyway, I get a blog for the GLPK. http://glpk.blogspot.com/. The one and the only one post is about how to set GLPK under VC 6.0 and Borland C++ 5.0.

Careers in the Decision, Risk and Management Sciences

2006-07-05T23:05:00.000-04:00

The article from science websiste discusses the career development.

The 10,000-member Institute for Operations Research and the Management Sciences
(INFORMS) has members working for homeland security, transportation, health
care, law enforcement, the military, and telecommunications. These people apply
scientific methods to help improve decision-making, management, and operations
for many different kinds of organizations, says Barry List, director of
marketing and public relations at INFORMS. "By using techniques such as
mathematical modeling to analyze complex situations, operations research
provides the power to make more effective decisions and build more productive
systems based on more complete data; consideration of all available options;
careful predictions of outcomes and estimates of risk; and the latest decision
tools and techniques

Usually, people in OR often thinks what people should do. Even under game theory, there is a strict restriction on the people behavior. So it would be fun to study what people would do under specific situation. Based on that, operations research analysists can conduct optimal response to the party they serves. Moreover, the strategy of avoiding the irrational impact from others can be developed.

Diaper and Beer

2006-06-07T22:40:00.000-04:00

From WSJ:

Tesco's computers often turn up counterintuitive results. Shoppers who buy diapers for the first time at a Tesco store can expect to receive coupons by mail for baby wipes, toys -- and beer. Tesco's analysis showed that new fathers tend to buy more beer because they are home with the baby and can't go to the pub.

Detailer can be founded in a blog called Data Doghouse.
Giving causation to the correlation given by the statistics is quite interesting:
More interesting and probable OR people can do is the following:
We have N items with correlation matrix, and we want to put them in the shelf togather to gain maximum profit. It is quite similar to the facility layout problem. And we may also hope to change the layout as little as we can, since the customer use to current layout.

Minority Report

2006-05-30T13:24:00.000-04:00

Is it true?

In Richmond, Va., police use predictive analysis to determine the
probability that a particular type of crime--armed robbery, auto theft,
murder--will occur in a specific area at a given time. Police lieutenants who
command the city's 12 sectors use desktop computers linked to the system to
decide where to deploy a mobile task force of 30 officers. "Based on the
predictive models, we deploy them almost every three or four hours," Police
Chief Rodney Monroe says.
Officers have arrested 16 fugitives and confiscated
18 guns based on the system's guidance. In the first week of May, Richmond had
no homicides, compared with three in the same week last year. Monroe attributes
that outcome, in part, to moving officers around based on the calculated
probability of shooting incidents. "It's more proactive," Monroe says. "We're
not waiting for a homicide to occur."

In the news, it shows the softwares apply several methods such as Markov decision process, stream mining, and support vector machines to predict the future. Also it is good news to inventory management.

IBM last week introduced an inventory management application for retailers
that uses built-in predictive analytics and replenishment rules to monitor
product inventory, develop safety stock, and recommend orders based on an
analysis of historical demand. The commercial app has been used by IBM
consultants for years.

Only 2.4% compound annual growth rate for SC planning

2006-05-30T12:08:00.000-04:00

Although the total market size for supply chain planning is already above 1 billion. But such low compounded annual growth rate indicates wave of SCM would be past soon.

The worldwide market for Supply Chain Planning is expected to grow at a
compounded annual growth rate (CAGR) of 2.4 percent over the next five
years.
The market was $1.05 billion in 2005 and is forecasted to be over
$1.18 billion
in 2010, according to a new ARC Advisory Group study, "Supply
Chain Planning
Worldwide Outlook: Market Forecast and Analysis Through
2010."

Nevertheless, I think the slow increase rate is due to lack of professionals. So companies cannot convert software to the profit.

Super Bowl of Operations Research

2006-05-15T21:32:00.000-04:00

Franz Edelman Award this year came to Warner Robins Air Logistics Center. They use "Critical Chain" to reduce repair and overhaul in the depot from thirteen to seven and repaired time drop 33%. When I try to find a little more detail from www.Informs.org, unfortunately there is even no further information. As I think "Critical Chain" is pretty similar to critical path. Maybe it considers undergoing tasks through repair network instead of static analysis of critical path and adjust resource correspondingly. I am not sure whether it will consider stochastical demand and repair time. If it doesn't invovle stochastical element in their model, it still has much space to improve.

If I get a chance to join this competition, I would like to implement my new idea about advertisement in TV.

Currently, the Ad during the program is quite even distributed. If I was watching a program and there is Ad, I would change the channel. I may find more attractive program or even attractive. Even if I cannot find one, I may switch back to the current channel without watching much Ad. Suppose now we use nearly exponential distributed Ad news. I think it may prevent people sticking changing channel.

First they cannot clear idea how long the Ad will be so they will change back quicker even if they change back

Second since it is exponential distributed, when you are back you will still see same expected length of Ad. given it is still Ad time

Third since there is more time each person watching Ad, the company can reduce total Ad length. It brings more satisfication of customer.

Ph.D. Thesis Research: Where do I Start?

2006-05-08T22:47:00.000-04:00

I just finish reading this article. It is worth reading. The main questions this article hope to address is indicated from the title. So which topic I should choose?

New and Important. Everyone knows it. But how to find such topic is still the art. Although this article is for Economy major. But it is applicable to other major. Also it is not limited to how to pick a topic, it also mentions other important success factors for a PhD student.

I post some section titles as well as my opinion:
1. How Do I Find ÂThe Right TopicÂ? (No right topic)
2. How do I know if I have an interesting topic? (Find a real world counterpart, discuss with others...)
3. Where do I start? Strategies for Research:
3.1 If you want to write applied theory, read empirics.
3.2 If you want to write empirics, read theory.
3.3 There is a ÂResearch FrontierÂ; Your job is to find it.
3.4 Go to weekly departmental seminars in your field. ( I like it and hope to ask a question every time)
3.5 Go to seminars of potential new assistant professors at your school. ( I will try this summer at IP seminar)
3.6 Read the working papers of the intellectual leaders in your narrowly focused research area. 3.7 Read the best journals selectively. ( Still have no time to read or at least scan MS, OR, MSOM regularly)
3.8 Talk, Talk, Talk! Write, Write, Write! ( Need to practice more )
3.9 Question Authority! ( That's what I like to do )
3.10 DonÂt Take Courses! ( I don't agree this. But it is true that I will choose to audit rather than taking)
3.11 DonÂt teach! ( The author must be a American who doesn't need to worry about teaching as a professor )
3.12 Dealing with advisors. ( Yeah. "the risks of saying nothing far outweigh the costs of occasionally saying something stupid (so long as you also occasionally say something interesting!).")
3.13 Your advisor is too nice!
3.14 Present your work whenever possible.
3.15 Consider writing your first paper jointly.
3.16 Writing matters. ( Still a long way to go )
3.17 Presentation matters. ( Is it easier than writting? )
3.18 Inspiration is where you find it

The semester is over and the summer is coming

2006-05-04T16:36:00.000-04:00

I finished all the final exams today. Now it is the time to set ambitious targets for this summer besides routine research work.

Three books to read
Zipkin: Foundations of Inventory Management
Ross: Stochastic Processes
Royden: Real Analysis

One course to audit
Derivatives and Risk Management

Sports
Running for half an hour three times a week
Doing Exercise at Gym for one hour two times a week
Tennis once a week

The benefit 'Dummy'

2006-04-25T22:57:00.000-04:00

Today's nonlinear class, I learned a trick by creating dummy variables and constrains to decompose the problem.
The original problem is

min f_1(x) - f_2(x)
st. x \in X_1 \cap X_2

Then we can transform it to

min f_1(y) - f_2(z)
st. y = z
y \in X_1 and z \in X_2

Then we can write the dual of transformed problem and do the decomposition.

Suppose f_1 and f_2 is convex.
We imagine the geometric meaning, for each Lagrange vector for the constrain y = z. We can find the corresponding tangent lines of the f_1 and -f_2. If their x happens to be the same, then we done. Otherwise we update Lagrange vectors. The way to update Lagrange vector is based on sign of difference of two tangent points.

Operational statistics: Do things togather rather than separately

2006-04-22T23:55:00.000-04:00

This friday Max Shen came to give talk on the 'operational statistics'. The idea is that when you seperate forecast and decision, you lose the potential saving. For example, if we don't know the demand distribution exactly under newsboy setting, we will forecast the parameters and make the order decision. It gives us lower profit. But if you directly design order quantity function instead of predict demand parameters and transform it to order quantity, we can gain a lot. It's managerial insight behind this paper to do things togather than separately.

But I just wonder what makes it work mathematically.

At seminar, I ask the question about unbaised and baised paramters. After I think it a little, I believe, by transform their order function into sample estimation space. It is likely that we get the baised estimation but with smaller variance. Then there is map between sample estimation space to the revenue space. Now we have two kinds of distribution of estimation. One is unbaised with large variance and the other one is baised with smaller variance. Suppose the map is concave, you can image the latter one will have bigger revenue if that estimation is well designed. So from statistical perspective, we need to balance how bais and how big variance of your estimation from the point of map between estimation and your objective function.

It is still hard for me to express these subtle things real time. Usually I have a very rough thought and need a discussion to clarify what I really think finally. Messy minds.

My computer restart when simulated multistage scenario tree program runs for three days

2006-04-19T20:34:00.000-04:00

I didn't anticipate that this simulation takes such long time before. Originally I think it may take just 24 hours. I try 100 times smaller sample size, it takes only 15 mins. After running for one days, I realize that simulation time isn't linear on the sample size in this problem. Because I create a linked list to compare the scenario it already creates, then comparison complexity is more close to O(n^2) not O(n) at first 1 or 2 period.

Next time I need to consider more before running big problem and store the intermediate result.

But it is so strange that my computer will restart itself. Maybe it is due to Microsoft automatical update.

close loop and open loop

2006-04-18T11:06:00.000-04:00

Yesterday, I read an article. The system described seems to me as close loop system. But instead it is called open loop system. So I search google and wikipedia today. Now I am total confused.
http://www.atp.ruhr-uni-bochum.de/rt1/syscontrol/node4.html
As I understand, for the open loop the state used to adjust control is variable outside the system. That is, if the system is disturbed, the control won't response.

But in http://en.wikipedia.org/wiki/Open-loop_controller
It defines

An open-loop controller is a type of reactive
controller which computes its input into the system using only the current
state and its model of the system. It does not use feedback to determine if its
input has achieved the desired goal.

By this definition, current state can be affected by disturbance. That is, if the system is disturbed, the control may response later instead of immediately.

Which understanding is right?

My comments on Getting More from Call Centers

2006-04-16T11:08:00.000-04:00

1. Set priority Q. Use routing software to send different type of customer with different tast to the corresponding agent. The problem is what type queue it will be V, M, N or others? If it is combined with time variant demand, how to set the parameters.

2. Agent learning. In the article, it said

Ideally, supervisors should spend 70 percent of their time coaching, and the
number of agents they monitor should reflect the team's role, so that, say, a
general-service queue would be supported by a 1:18 coach-to-agent ratio, while a
vital sales or support queue that needed more coaching would enjoy a 1:14 ratio.
Coaches should combine side-by-side training with remote listening (in which the
agent is unaware of being monitored) and should provide immediate feedback in
both cases. The coach is also responsible for sharing best practices with
agents.

The ratio is surprisingly low. Is this ratio indeed reasonable?

3. How to evaluate quality decreasing via outsourcing. What is the relationship between quality and reveneue

4. What performance measures are effective. Since I have do several projects related to the PM, it is one of most interests to me. The performance measures mentioned in the article are total revenue per month, average handling time,... It also mentions that without good design, the effect of call churning will happen. Like balance score card, it is crucial to know how to convert the individual targets to the overall score which is related to the incentive.

I never see any research paper on the part 4 before. Maybe human resource research will be more relavent than our OR field. But it is important issue.

Good Article!