Hobby Projects

github links to be posted

Playing 2048 via Reinforcement Learning

2048 is a highly addictive game, at least for the first few times. While it seems simple for me and my friends to grasp a good strategy after a few trials and errors, the strategy requires fairly long-term planning. I am curious whether and how techniques used in AlphaGo will find the right strategy for this seemingly simple game.

Reverse Engineering Authors from Their “Work”

Artists, writers, engineers all create their masterpiece in unique ways, although sometimes the uniqueness is not easy to notice. This is an exercise to see if deep learning can identify the authors from their drawings, articles and software codes. And if the answer is yes, then can the machine learn the unique features/signatures.

Markdown to PPTX Converter

During work, I spent too much time copy-pasting figures, tables and equations to PowerPoints, not to mention the time “wasted” for making beautifully resized and aligned images with consistent linewidths and arrowtypes. While something like beamer would work for academics, PowerPoint is sometimes inevitable. With this handy tool, I can now generate PPT reports by writing scripts or markdown language.