I currently run Arch Linux in Virtual Box on my MS Surface Pro 3. I must say, this is the most efficient, fluid system I have ever owned, and the keyboard is superb. This page serves as a reference both for me and for anyone who wants to set up a similar system. It also includes useful scripts for automating various common tasks.
Machine Learning Course
As a junior in high school, I struggled through the basics of machine learning--I lacked the mathematical background to understand most university level resources, and there were few to no resources for high school students. That changes now. Hundreds of hours have gone into this. I am not a professor, and I do not recommend that you make this your only resource, but it serves as a great companion to Prof. Andrew Ng.'s (one of my greatest heroes) Coursera course, for those of you having difficulty. All but the CNN code is on Github. Recommended math level: single variable calculus with a willingness to learn partial differentiation. Recommended language: Matlab.
Water Cooled Desktop
I swapped my laptop for a tablet, but that meant giving up a lot of processing power and memory that I need for running large simulations. This was the overkill solution: overclockable past 4.6 GHz while maintaining temperatures lower than my tablet's, thanks to the custom liquid cooling loop. I now run most of my intensive code on this over SSH.
Research Internship: Visual Genome
I spent the summer of 2014 working in Stanford's Computer Vision Lab with Justin Johnson on his project, Visual Genome. While my involvement comprised several sub-projects (mostly prototyping machine learning pipelines) and potentially novel approaches, I am withholding this information until I receive permission to release it.
Feed Forward Artificial Neural Network
Currently my favorite algorithm. In all of computer science. Why? Because it is, to my knowledge, the most general, elegant graphical approach to machine learning that remains just within the reach of even high school students, provided they are sufficiently passionate. Implementations in python and Matlab. First time students trying to understand this algorithm: see my miniature machine learning course.
Virtual Reality Glove
Allows the user to map arbitrary gestures to arbitrary keystrokes. This was a three day project that gave me the opportunity to combine electronics, modular design, and machine learning, all while passing data through several programming languages. Keypress emulation is limited to OSX but is easily extensible.
An easy CLI python image editor. Simple, quick efficient. The image shown was generated in only a minute or two of experimentation via:
python pImg.py -c rgbNoise 1920 1080 -c brightness -80 -c rgbScale 0.5 1.2 2.0 -c rgbShift -10 20 30 -c fade b k 0 1 False -c fade l r 0 100 -c fade r g 0 80 -c contrast 1.5 -c brightness -30 -c rgbScale 1 1 1.8
A very old project, but it comes with pretty visuals! This was my second step into machine learning, and this time I included a script for drawing decision boundaries. This page is mainly a link to my Github. First time students should check out my miniature machine learning course.
Touch based MP3 prototype. All controls can easily be manipulated without visual aid. In a smaller form, optimal for joggers etc. (e.g. facilitates volume, track, loop, and shuffle control).
A very old MATLAB project with Tiago Lima. My first MATLAB project, in fact. Cleverly combines several (extremely) basic operations in an attempt to compute the total value of coins in an image. Surprisingly, it works fairly well when given a clean image.
Generic, vectorized implementation of linear regression in n variables via gradient descent. Included for nostalgic purposes, as it was my first step into the world of machine learning. Python and Matlab implementations.