Project Checkpoint 1

For this assignment, you should make an appointment with the GSI overseeing your team project. You should receive an email from them with details about how to sign up for an appointment during the week of 3/11/19. The goals of this checkpoint are to make sure you have the resources in place to move forward on your project.

Dataset

If your project uses a particular dataset, you should demonstrate that you have already acquired all the data that you need. You should also demonstrate that you have done any required pre-processing (e.g. tokenizing, dictionary-building for text data, clipping/format conversion for images, computed image sequences from video). If you havent done this yet, you should demonstrate that you have the resources to do it in a reasonable amount of time, e.g. by preprocessing a sample of the dataset and benchmarking to ensure you can complete processing of the full dataset.

Simulators/Games

If your project is based on a simulator or game, make sure you have an usable API into the game. This is easy for many popular RL environments (Atari, Vizdoom, Dota, Torq, Grand Theft Auto) but may be arbitrarily complicated for environments without an existing API. Ideally you should write a short script to demonstrate that you're able to run a random agent in the environment.

Computing Resources

You should verify that you have the computing resources needed to train the model you're developing on the dataset or environment that you;ve chosen. You'll often have to do that by estimating from benchmarks from other people's systems. As part of this, design your model to allow a wide range of complexity, i.e. dimensions of latent variables, number of layers etc. Consider shrinking or down-sampling your data if appropriate.

Checkin

There is no online submission for this assignment. You will receive credit from the GSI when you have attended a meeting and verified that your project is ready to go.