This post is based on the material I prepare for a learning session in my department. I have also made a slide and recorded video, but I cannot share them for corporate reasons. Compared with some classic machine learning tasks, reinforcement learning (RL) is not easy to learn. It’s very…

Recently, I read an article [1] about selection bias in online a/b tests, which describes a simple but intriguing phenomenon, i.e., app features usually produce less gain in production than in testing. I found this problem very interesting and decide to dive into it. An Example Intuitive Explanation Hypothesis Test…

Recently, I have watched a lecture on youtube about the practice of writing a good research paper. I find this video very useful; here I briefly summarize the lecture in this post. Why Write Papers How to Write a Paper ⠀⠀⠀ General Principle ⠀⠀⠀ Recommended Structure ⠀⠀⠀ Abstract ⠀⠀⠀…

In Post I and II, I have introduced some basics of AI game-playing algorithms, and Monte Carlo Tree Search (MCTS) [1], a popular algorithm suitable for solving perfect information games. In Part III, which is this post, I will introduce Counterfactual Regret Minimization (CFR) [2], a popular algorithm for solving…

In the last post, I introduce some background knowledge for game ai, namely, game representation, game categorization, and three important game-solving concepts. In Part II, which is this post, I will introduce Monte Carlo Tree Search (MCTS), a popular algorithm suitable for solving perfect information games. Specifically, I will explain …

Game AI has always been a popular field with many impressive achievements. Next, I will start writing a series of three posts, talking about some fundamental and important game-playing algorithms. Part I — Game Basics Part II — Monte Carlo Tree Search (MCTS) Part III — Counterfactual Regret Minimization (CFR) …

· Introduction · Key Questions to Understand LDA ∘ Why Dirichlet Distribution? ∘ Why is Exact Inference intractable? ∘ How Variational Inference makes it tractable? · Summary · Reference In this post, we will look at the Latent Dirichlet Allocation (LDA). LDA was proposed at [1] in 2003 and was…

In this post, we will look at the Neural Process (NP), a model that borrows the concepts from Gaussian Process (GP) and Neural Network (NN). The Neural Process was proposed in the paper Neural Processes. …

Rigging the Lottery: Making All Tickets Winners [Supplemental] ICML 2020 This paper proposes a new method to train a sparse neural network, which raises the theoretical limit of a sparse neural network’s trainable size under memory constraint and yields higher accuracy. …