Skip to content

Latest commit

 

History

History
35 lines (24 loc) · 947 Bytes

README.md

File metadata and controls

35 lines (24 loc) · 947 Bytes

Probabilistic Programming 6998 Final Project

Our project extends the ideas of Noise Contrastive Priors to the classification setting. It is based on:

Noise Contrastive Priors

This project contains the source code for Noise Contrastive Priors (NCP) by Danijar Hafner, Dustin Tran, Alex Irpan, Timothy Lillicrap, James Davidson.

Running the code

Install dependencies:

pip3 install numpy tensorflow tensorflow_probability matplotlib ruamel.yaml

Active learning on the toy data set:

python3 -m ncp.scripts.toy_active --seeds 20 --logdir /path/to/logdir

Active learning on the flights delay data set:

python3 -m ncp.scripts.flights_active \
  --seeds 10 \
  --dataset /path/to/dataset \
  --logdir /path/to/logdir

The data set path should be a file prefix for four Numpy files named <dataset>-train-inputs.npy, <dataset>-train-targets.npy, <dataset>-test-inputs.npy, <dataset>-test-targets.npy.