College Life at Tsinghua University, Beijing. 在清华的生活

Emergent behaviors are pervasive in Physics, Chemistry and other natural sciences. For example, when snowflakes fall to the ground, its shape changes due to intrinsic complicated interactions under different thermal and humid conditions.

One of the very interesting experiments concerning emergent behaviors is John Conway’s Game of Life. In this game, a set of simple rules are pre-defined for each unit of the map. At each step, a unit interacts with its nearby units. As time goes, complex patterns and behaviors can be observed.

Stephen Wolfram gave a systematic review of simple rules to generate complex patterns, which he mentioned as cellular automata. Cellular automata was also used to explain the formation of snow flakes and applied to many applications like image processing. However, as Ray Kurzweil pointed out in his book “Singularity is Near”, the patterns generated by the cellular automata is limited in that no patterns more complex than those depicted in Wolfram’s book are observed. For example, no recognizable images, no trees, no humans can be produced through Wolfram’s approach.

From the informational theoretic point of view, complex patterns contain much more information. However, if these information can be generated through simple rules in an emergent manner, then much part of this information is redundant. As a result, these rules can be summarized by observing the patterns through machine learning, just as we human beings learn by looking at our daily experiences.

I tried machine learning on generating cellular automata from a complex pattern, say “Chrome logo”.

 

This can be achieved if we consider each node in the map as a computing node, say a neuron in the Neural Network. Then we can use the Back Propagation algorithm to learn the model by applying a critic at each of the pixels at the image. At first, no such constraints like “rules should be as simple as possible” is used. So the weight of the neural network is free. We start with the initial value:

 

Through several iterations, we ended up with a Neural Network output like this:

 

 

However, this result is not satisfactory since our automata does not generate complex patterns through simple rules. Instead, its rules are quite complicated. In fact, much of the information is stored right in the weights of the neural network.

In order to restrict the rules the automata generated can use, we can put restrictions on the Neural Network by reducing its receptive field and synchronizing the value each weight can take. By doing this, I find the learning algorithm  fails to fit the image — a high bias exists. This probably explains why no more complex patterns are observed in Wolfram’s book — no models exists with a low bias.

In the future, I hope I can apply theories of machine learning to prove the limitation of Wolfram’s cellular automata.

 

A brief summary is made of the mathematical modeling experience with CPCCWI Students at Nanjing on August 3rd, 2012. 

 

The problem to solve is to predict the gold medal winner at Olympic London 2012, in terms of Tennis, Basketball, Diving and Hurdle.

 

At first sight, Tennis and Basketball, as team-based games, are similar. So are Diving and Hurdle. However, there are differences due to limitation in data. We had much data than expected in Tennis because the players have competed with each other for years and their strength and weakness are known. On the other hand, we couldn’t have much data for Basketball since we noticed teams like U.S. basketball team have completely new compositions which have not played much.

 

The solution of Tennis was straight forward if we could statistically calculate the probability of one player winning another. Then follows the probability for each winner to be the champion. Due to the limitation in basketball data (only two rounds with two groups each round have been played), we turned to finding out the parameters (features) that determined the strength of a basketball team. This led to machine learning the data, which was done by a Neural Network of 12 hidden nodes handling features like PPG(points per game), 3P%(rate of 3-point shoots), etc. The result turned out quite good, predicting the U.S. basketball team winning with high probability(91%).

 

When it comes to games like diving and hurdles, machine learning with neural networks might not work since too many unknown factors exist in the process of the game. To model that uncertainty, we assumed each player’s performance follow a gaussian distribution, with which the ensemble ranking of players could be calculated by method of Monte-Carlo. This predicted two chinese diving players would win at a probability around 30% respectively for diving. For 110 men hurdle, Robles win at a probability around 40%, slightly higher than Liu Xiang.

 

The neural network model and the monte-carlo calculation dawned on me almost unexpectedly as gut feelings, but turned out to be quite good methods. This proved the importance of knowing models and using them with flexibility. 

上个学期申请转去姚班的申请信。今天正好不小心翻出来了,就贴在这里吧。

 

Transfer Application

Tianlin Shi

 

“Have the courage to follow your heart and intuition,” said Steve Jobs at the 2005 Stanford Commencement Address. For years I’ve been obsessed by Science and its huge impact on life. Physics, where relativity and quantum theory occurred as great discoveries in the 20th century, was once the great science in my heart and led me to the door of Dept. of Physics, Tsinghua University. However, physics in the 21st century seems less exiting to me than it was while another great science is in the making, the science that is shaping the information age, creating myriad incredible interdisciplinary areas and producing countless challenging unsolved problems. That is Computer Science. I’m applying to IIIS, China’s most pioneering institute in computer science, for transfer and future study. This application essay brings an introduction to me and more detailed reasons of such a transfer.

 

When I was an epsilon, I made myself the uncommon English name “Strin”, which is homophonic to my Chinese name. Strin is the string without ‘g’, while the latter has deep connection with music and later coincidentally with computer science. In the childhood, exposure to activities like collecting stones and reading books instilled in me the interest and curiosity about science. There was a time when an isolated mathematical puzzle “the Black Hole Number 6174 Problem” came into my eye from one of the science books, which was claimed to be unsolved. But I tried to solve it. It was several days of hard thinking for me before finally a proof of the reduced version “495” struck on my intrigued mind. This was the first time ever that I was so lucky not as a scientist but as a problem-solving novice to be able to experience the joy of science and how human knowledge can be achieved with efforts and creativity.

 

Unlike most students who went to Math Olympiad or struggled for College Entrance Exams, my stories were totally different at Shanghai High School, where I had the fortune to have the very first taste of the Science Education Program. Aimed at the early cultivation of science geniuses, the program showed me the most spectacular scenery of the science world with specially designed courses, empowered a mind to solve realistic problems with compelling lectures and plotted the edge-cutting researches with mentors from Shanghai JiaoTong University. As a student, I helped do optical experiments with the Physics professors from whom I learned to focus and to be careful. Alongside with physics, the initial exposure to computer science had also been possible thanks to professors from the Dept. of Information Security. I devised algorithms that detected abnormal network processes from the patterns of packet and connection activities. I wrote programs that accurately recognized voice during the phone conversation and turned it into real-time translation. From my point of view, the consequence of such experiences is not the honor of winning at Intel Science Fair, but an ignited passion and curiosity of science and nature that will never vanish.

 

One more noteworthy experience in high school was the exchange study at Crystal Spring Uplands, San Francisco. What impressed me most was not the enjoyable weather, the delicious California food or the blue sea at Fisherman’s wharf, but the encouraging atmosphere and the exciting lectures at Stanford, Berkeley and the Silicon Valley. I marveled at the large printing machine at HP, the revolutionary cloud computing technologies at SUN, the mass production of chips at Intel, etc. The lesson I learned was that science and technologies are really changing the world at a tremendous and unimaginable speed.

 

I transfer for three reasons. First of all, I appreciate challenges in study and Yao Class features courses that get me interested and stimulated. Last semester, I went to a course in Artificial Intelligence at the computer science department, which combined challenging problem sets from Berkeley AI project. One of the problems was to define a heuristic function for Pac-Man and the best result expanded more than 7000 nodes. I pondered over this problem with effort for a few days and eventually came up with a result that expanded around 200 nodes and that really impressed the teacher. The story actually tells me that only by stimulating and challenging myself can I constantly make myself uplifted by a tiny bit.

 

Secondly, I enjoy learning from others and in college I believe this form of learning is just as important as learning from books. Students at Yao Class are all from different backgrounds and have the same strong passion for Computer Science. It’s fun and beneficial to discuss problems and share ideas with them when attending a class, having a meal or even during the sleep. Such kinds of communication will branch out even further with frequently available international exchanges.

 

Last but not the least, Yao Class is more research oriented while I just developed great interest in Quantum Information and Quantum Computation, for which I spent much of my spare time reading books, news and papers. Having finished the book by Nielson, Chuang as well as the course by Prof.Long, I’m convinced that this is the new land of computing to fertilize next revolutionary areas of computer science, involving many questions unanswered and more ideas undeveloped. The Quantum Information Center at IIIS, which gathers so many fascinating projects to work on and so many sophisticated researchers to chat with, can be the right place to start my journey of the great science world.

 

2011.04.17

Tsinghua

 

 

 

 

We are immersed in technology. “Nomophobia”, the pathological fear of being away from one’s mobile phone, was unimaginable even ten years ago. Technologies, like water and air, have become a indispensable resource we human beings live on.

Technologies are changing the way we perceive the world. Google Glass extends the vision of our eyes by enabling instant “Snap-shots”. Satellite map of the city, once only accessible to the Security Agency, is now available to everyone, via laptops, tablets and smartphones.

But what is changed is not only how we perceive the world, but what we see as the world. Our perspectives are shaped by technology. One clinical study by famous sociologist Sherry Turkle at MIT is to observe the reaction of the old companioned by robots. What she found is that the old people are made happy, despite the awareness of robots being machines.

Many people will describe technologies, say computers, just as tools. The word “just” is actually quite deceiving, behind which is a incredible fact that technology is shaping us, thoroughly. Lumosity, started in 2007, is an online brain training program that helps improve mental intelligence by integrating research findings in Psychology and Neuroplasticity. The results are remarkable and greatly confirm the revolutionary role of technology.

The singularity, the moment machines surpass human intelligence, is near. The period from now to that singularity is what Parag Khanna and Ayesha Khanna in their TED book “Hybrid Reality” call “Human-Tech coevolution”. We are indeed co-evolving with technology and distorting the theory of evolution conceived by Charles Darwin. The nearer the singularity, the greater the gravitational force. And we shall all be ready for that.

Why Blogging?

I used to own a blog in high school( hi.baidu.com/ohstrin ). And I think it necessary to start a new blog in college.

Almost one week ago, Facebook, the social network giant, went public. People may ask why use such kind of old-fashioned Internet tool as Blog. But a blog is totally different from social networks. Social networks like twitter and facebook usually brings too much information. It’s a bit “noisy”, while I would rather choose a place as quiet as “wordpress” to share my ideas and write down experiences.

College is different from High School, which means the same for blogging. Let me state the aim of this blog:

Aim of this Blog:

(a) Share my experience in study, research and social activities.

(b) Let people know China, especially the higher education here at Tsinghua university.

(c) Practicing writing should be part of the life.

Remember, this blog is not only about me. (PS: about me)

See this at docs.google.com

Tag Cloud

Follow

Get every new post delivered to your Inbox.