I haven’t updated this blog since I got busy at work. However, I have <ahem> found myself with more free time lately, so I finally have had a chance to do something I’ve been thinking about for a long time. I’ve taken the inverse limit visualizer I translated to Python 2 years ago and made it an interactive dashboard. You can modify the bonding map by adding/dragging points around and see the changes to the inverse limit in real time. Try it out here! Be patient while it loads, it takes a few moments. Also be aware that the server that it runs on will shut down an instance after 20 minutes of inactivity.

In keeping with the theme of autoencoders, in this notebook I perform an experiment with an autoencoder to resolve a problem I had in training a GAN.

I’ve added a project I did over the past few days.  It’s an implementation of an autoencoder using fast.ai version 1.0.  Autoencoders are an interesting machine learning exercise that can be used on any data set for unsupervised pretraining, feature extraction, and other things.

Check it out here!

In grad school a few years ago I was exploring a topic in mathematics called inverse limits.  Inverse limits are a type of “space” that are constructed in a specific way.  I wanted a way to quickly visualize these spaces without working out a representation by hand, so I wrote a visualizer in MatLab.  More recently, I ported the code to Python and cleaned it up (a lot).  You can see the results here.

I’ve had the idea for a while to create a hub to express the things I’m learning about data science and deep learning, interesting things about mathematics, or anything else that comes to mind.  I also plan to showcase past and present projects here.