Hi!

I'm a grad student studying ML



Baja - Part 2

Controlling a Robot

In a previous post, I hacked an RC car chassis to create a little robot. This post is to introduce RTCBot, a Python library to remote-control the robot over wifi and 4g. It uses WebRTC to stream the robot’s video from a webcam to the browser, and send back controls from a keyboard or an xbox controller.

Try it here





The Great Filter

A doomsday discovery

Thus far, astronomical observation has not shown evidence of extraterrestrial civilizations. This leads us to the famous question: Where is everybody? This apparent lack of life outside Earth seems to be at odds with the... continue »

Gun Laws

I have recently seen quite a bit of outrage on social media aimed towards gun control in the USA. I don’t know why facebook’s algorithms decided that I should care about a topic that bears... continue »

Visualizing Causal Models

I found it pretty annoying to quickly generate pictures of causal models in my research, for inclusion in presentations and latex documents. To fix this problem, I’ve used the visjs library to generate pngs of publication-quality causal models.

The code, as always, is on github

Try it here



An Update

In middle 2014 I started thinking about a data-gathering tool for quantified-self info, with the explicit purpose of treating life as a machine learning problem. I am happy to say that for the past year, Joseph Lewis and I have been working on something I think is quite incredible.

We are very close to having a first version of the software we wrote available. I just wanted to say that I can’t wait to show off what we have accomplished. Our hope is to have a nice demo running by January!


Milestone 1

In the last week of September I started on a very interesting personal project. I was finished with homework for the week, and decided that it couldn’t hurt to play around with an idea that was bothering me for a while.

“It should take me no more than a couple days to get something up and running”, I thought.

One month later…

Nothing is up and running yet.

For the past 4 weeks I have spent every second of my free time (and some not-so-free time stolen from other obligations) to work on this project.

It wasn’t all for nothing, though. I have finally reached the first milestone. What does that mean? Well, I have the code that underlies the thing that will underlie the actual code that does stuff. In effect, I have finished a very convoluted and complex boilerplate. So about 500 more milestones, and I might have the equivalent of “hello world”.

Nevertheless, I am very proud of myself, despite not having much to show.

Now it is time for a day or two of break, so that I can catch up with my obligations, and then on to milestone 2!


The Great Balancing Act

There are several methods for reading papers. There is the “read-on-a-laptop” method, which gives a good idea of what is going on, and then there is the “print-the-paper-and-spend-6-hours-going-through-it-word-by-word” method. The second method seems to inevitably... continue »


The king is dead, long live the king!

The time has come for me to say goodbye to the Kwiat Group, whose members have for the past 3 years been incessantly a source of friendship and inspiration. It was there that I first tasted the joys of research. I am extremely grateful to them for helping me discover one of the greatest sources of joy and satisfaction in my life. I will definitely bring them many cookies next time I am in town.

In leaving Kwiat Group, I am leaving the world of Physics research, and moving on to Machine Learning. I’m spending the summer working with Dan Gauthier’s group at Duke University, focusing on Reservoir Computing.

Imagine throwing a pebble into a pond which has some nonlinear properties. It turns out that it might be possible to more easily extract information about the pebble’s properties through the excitations of the waves and movements of the pond upon the pebble’s impact, than by looking at the original pebble! These Liquid State Networks, which are called Reservoir Computers when in the form of a Neural Network, can achieve good performance in machine learning tasks, despite having only one hidden layer, which is both generated randomly and fixed!

Sounds awesome, so I will see where this track of research leads me.


New Website

Thanks to Rebecca from Kwiat Group, I discovered Jekyll, which allows for quick generation of static websites.

I don’t want to use Drupal, or any PHP-heavy tools for my website, since my VPS has 128MB of RAM. Jekyll is a step up from manually editing HTML files, while still giving me full control of a lightweight static website.

One of the cool things is that it allows writing content in markdown, which automatically does syntax highlighting with pygments and math with MathJax.

import antigravity
print "Hello World!"

Pretty cool, if you ask me.


TDC Open Sourced!

We have open-sourced the code used for time tagger control. I am super proud of it! Check it out here: Kwiat Group Github


Electric Bike Trip

What better way to finish college than by biking from Illinois to California? I don’t know - and that’s why Spencer Gore and I are doing it!

Update: The trip was fun, but we ended up in Kansas City instead of California.


Android Development

I have begun learning Android development. My first goal is to create a program which can recreate my schedule from GPS coordinates and accelerometer data. I want to be able to know exactly how many times I got out of my desk, how many times I picked up my phone to talk to friends, and so on.

This program will eat my battery alive (accelerometer needs wakelock), but I think that knowing what I actually spend my time on is worth carrying a charger around for a week or two.


TDC Documentation Finished

Part of my work for Kwiat group involves interfacing with time-to-digital converters (TDCs). This is especially important, since my main job is to develop Information Reconciliation codes for use in our higher-dimensional quantum cryptography project. These codes require a very fast way to interface with multiple TDCs.

To that end, I created a C library which enables a standardized method of access to time stamp data. I am proud to say that library is currently used on several research projects within Kwiat group, and is interfaced with 3 different models of time tagger.

So the news-worthy update right now is that I have finished a rough draft of the documentation for this library, which can hopefully be understandable to students without much programming knowledge. You can see it here!

I will put a link to the code once it shows up on the Kwiat group’s website.


Send People your Files

So you have a large file. You want to send it to someone. You can use Netcat!

#!/bin/bash
{ echo -ne "HTTP/1.0 200 OK\r\nContent-Type: $(file -b --mime-type "$2")\r\nContent-Length: $(stat -c%s "$2")\r\n\r\n"; cat "$2"; } | nc -l "$1"

This script is based on wikipedia’s netcat example. It is updated to include some information which allows streaming videos (VLC wants a file type and length).

To make the file ‘file.mp4’ available on port 1080, you run the bash script:

./fileserver 1080 file.mp4

Once your friend downloads it, the script returns - it only makes the file available once.

When you run the script, all the other person needs to do is point their web browser (or VLC/mplayer) at your ip and port number.


Watch youtube on the Raspberry Pi

The raspberry pi, however awesome, does not have the power to run youtube well in the default distro. Even if you don’t have X installed, a simple script allows you to take a youtube video’s URL and instantly start streaming its video!

To use this, you will have to get a working omxplayer and the awesome youtube-dl library. If you are using debian, they should be available using apt-get.

#!/bin/bash
#Different formats are: 18-360p, 22-720p, 37-1080p
omxplayer -o hdmi $(youtube-dl -g "$1") #Standard def
#omxplayer -o hdmi $(youtube-dl -f 22 -g "$1") #720
#omxplayer -o hdmi $(youtube-dl -f 37 -g "$1") #1080

Voila, instant streaming youtube videos!


nDB



Nova


DXPy