Skip to content
Source files for reviewing a paper from ICLR 2019
Python
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
README.md
generate_tables.py
heatmap.py
helperFunctions.py
neuralNetwork.py
neuralNetworkCIFAR.py
run.py

README.md

ICLR Reproducibility Challenge 2019

Reproducibility Report: Issue on Github

Original Paper: OpenReview

Our submission: OpenReview

The CNN for MNIST is described in neuralNetwork.py, and for CIFAR 10 in neuralNetworkCIFAR.py.

Running Experiments (MNIST)

The experiments can be run by: python run.py [--optimizer <name>] [--rerun count] [--epoch count]

Results are stored in data/experiment_result.pickle.

Visualizing results

heatmap.py is a script that opens the data (created after running run.py) data/experiment_result.pickle and creates the heatmaps found in the report

generate_tables.py is a script that opens the data (created after running run.py) data/experiment_result.pickle and generates the accuracy table found in the report

You can’t perform that action at this time.