For the past couple years, I have worked as a teaching assistant for UVa’s CS 2150 (Program and Data Representation) course. We recently started a GroupMe chat for the course staff, and I thought it would be fun to create a chatbot to help remind all the TAs to submit timesheets, keep track of when people are holding office hours, and remember when/where TA meetings are being held. Setting up a basic chatbot is a lot simpler than it sounds and is really fun–I wrote my bot from scratch using Python in just one day.

The 2150 chatbot

GroupMe Bot Overview

GroupMe has a very brief tutorial explaining how their API may be used for bots. The easiest way to create a bot is through their web form, which prompts you for the bot’s name, callback URL (technically optional, but you want it), avatar URL (optional), and the name of the group where the bot will live. Once you’ve done this, you will be given a unique bot ID token. Anyone with this token can pretend to be your bot, so keep it safe. (Security is somewhat laughable here: your bot asserts its ID and the server believes it with no “login” procedure.) We’ll talk more about the callback URL in a moment; for now, just leave it blank.

Once you’ve done these steps, you have created a bot–as far as GroupMe is concerned. If you send a specifically formatted JSON mssage, the newly created bot will post in your group. However, if left at this point, your “bot” is little more than a proxy for human-written messages submitted with curl. Your bot needs some way of reading messages sent to the group, formulating a response, and only then sending messages to the GroupMe servers.

While interning at Microsoft over the summer, I received a first-generation Surface Book with an i5-6300U CPU (2.4 GHz dual core with up to 3.0 GHz), 8GB RAM, and a “GeForce GPU” (officially unnamed, but believed to be equivalent to a GT 940). This is a huge step up from my older laptop, so I wanted to set it up for my ML work. In this post, I’ll outline how I set it up with TensorFlow and GPU acceleration.

CUDA + cuDNN

If you want to use GPU acceleration, the typical way to do so is with NVIDIA’s CUDA API. CUDA 8.0 is compatible with the Surface Book and is (as of this writing) the most up-to-date version of CUDA. Download it from the NVIDIA website and run their installer.

For work with deep learning, you’ll also want to install cuDNN. To install, just download the library from NVIDIA’s website and unzip it in a convenient place (I chose C:\cudnn). The only “installation” you need to do is to add C:\cudnn\bin to your PATH environment variable.

Nearly everyone is familiar with two-dimensional plots, and most college students in the hard sciences are familiar with three dimensional plots. However, modern datasets are rarely two- or three-dimensional. In machine learning, it is commonplace to have dozens if not hundreds of dimensions, and even human-generated datasets can have a dozen or so dimensions. At the same time, visualization is an important first step in working with data. In this blog entry, I’ll explore how we can use Python to work with n-dimensional data, where $n\geq 4$.

Packages

I’m going to assume we have the numpy, pandas, matplotlib, and sklearn packages installed for Python. In particular, the components I will use are as below:

1import matplotlib.pyplot as plt
2import pandas as pd
3
4from sklearn.decomposition import PCA as sklearnPCA
5from sklearn.discriminant_analysis import LinearDiscriminantAnalysis as LDA
6from sklearn.datasets.samples_generator import make_blobs
7
8from pandas.tools.plotting import parallel_coordinates

Plotting 2D Data

Before dealing with multidimensional data, let’s see how a scatter plot works with two-dimensional data in Python.

First, we’ll generate some random 2D data using sklearn.samples_generator.make_blobs. We’ll create three classes of points and plot each class in a different color. After running the following code, we have datapoints in X, while classifications are in y.

1X, y = make_blobs(n_samples=200, centers=3, n_features=2, random_state=0)

To create a 2D scatter plot, we simply use the scatter function from matplotlib. Since we want each class to be a separate color, we use the c parameter to set the datapoint color according to the y (class) vector.

As suggested by this post’s title, I spent this past summer as an intern with Microsoft in Redmond, Washington. The experience was highly educational for me–as my first (and last!) “real” internship, I learned a lot about software development and the importance of corporate culture, as well as discovering a lot about myself. Overall, the experience was a positive one, though, and I had an enormous amount of fun!

This is the first of a three-part series on my time at Microsoft. This post focuses on fun recreational activities for interns in the Seattle area.

Outdoors

Hiking in the North Cascades

The Pacific Northwest is home to some of the most amazing views I’ve ever seen. Seattle is conveniently located close to the beach, the mountains, Puget Sound, rainforests, and many hiking trails and campsites. Exploring the outdoors also has the advantage of being very inexpensive, which is great if you’re saving your internship money for college expenses. If you visit National Parks, consider the National Park Passport Program–if you’re going to once-in-a-lifetime parks, it’s a good idea to get your passbook stamped!

This summer, I’ve been working as an intern for Microsoft on the Direct2D/DirectWrite team. While I can’t really talk about what my work entails, I can talk about some of the fun things I’ve done this summer in my free time and the non-work-related components of my internship. I suppose most people wouldn’t start blogging about their internship by describing a bookstore, but I went to this place today and it was so incredible that I had to write about it.

In Capitol Hill, there’s a small store by the name Ada’s Technical Books. It’s in a house that’s been converted to a cafe and bookstore, and is quite possibly the most amazing bookstore I’ve ever seen. As you walk in, you’re greeted by an small cafe counter to your left and an open area to your right with short bookcases and comfy chairs. Toys, puzzles, and “Maker”-appropriate items like lockpicks and Raspberry Pis.

Who can resist a giant 555 timer?