On Proximal Policy Optimization

This is the story of a man named Proximal Policy Optimization (You may want to read this along with this article). His friends call him PPO. He's one of reinforcement learning most popular guys. PPO is a simple man, so simple that you can summarize how he acts in any given situation in just under …

The Summary of Basic Reinforcement Learning Algorithms

Vanilla Policy Gradient (VPG): Make trajectories leading to rewards more likely to be chosen by the policy On policy Discrete and continuous Trust Region Policy Optimization (TRPO): Make trajectories leading to rewards more likely to be chosen by the policy but take the largest step in each update such that the KL-divergence between the updated …

A refreshed take on AI

I recently came back to AI research with a new perspective. Previously, I built everything bottom up. Given that I wanted to do research in Computer Vision (CV), I would spend weeks learning all the basics of CV. It's like going into a rabbit hole, there are always so many things to learn that I …

The Deb: acceleration of free falls

The deb not surprisingly have a system of scientific laws, rather very basic ones. For example, they know that objects fall down equally fast regardless of their mass. However, the way that they came up with it is quite interesting. We human discovered that by dropping objects of different masses from the top of the …

The Deb: the garden corner

In the upper left corner of the garden, we have a wall that's made from lamps. Lots of lamps. Not normal lamps to be exact but fruits of the lightbulb tree. On the bottom of the wall, a tiny forest of mushroom grows. They give off beautiful dim green light at night. On the mushroom …

Illusion 100

We really want to have illusion 100 in Skyrim. But when we're confronted with the choice to use command cheat or not, we stopped. It's not about being a powerful mage. It's about becoming a powerful mage. If we can choose to suddenly have great ability in artificial intelligence, would we accept? We would but …

Working in AI

I'm proud to tell everyone that I'm an AI scientist. That I can make machine tells which picture is a hotdog and which is not.

What internship do you want?

We don’t know. We never tried to find out. We haven’t look deep enough. Internship is for learning specific, practical things, not to make profit for the company or to please people. So what specific, practical things do we want to learn? How do people do science, how do people invent things, how do people …

About summer internship

What’s great about it that you want it so much? It’s a good thing to put on resume to make hirers want me. What’s so great about having a job? So that I won’t be like the arrogant jobless people that my family so fond of talking about. So, if you get a job, who …

So what if the worst thing happen?

Suppose after finishing college we can't find a job in America. Is that so bad? We don't know. If we don't know then it must not be bad. We don't even know if we want it or not. What do we want to do after we graduate? No, why wait until graduation. We'll do it …