This is a playground for verifying results of and experimenting with MIMO lightweight ensembles.
reference
directory contains reference implementation of MIMO as described in the study.
custom
directory contains custom code for testing.
run_examples.sh
script is an entry point for testing MIMO with different combinations of weight, depth and ensemble size. chmod +x run_examples.sh
before running.
This directory contains the original scripts used for the paper's benchmark results, as well as experiments analyzing MIMO.
NOTE: An updated version of the codebase, further expanding on the original results, can be found in Uncertainty Baselines.
ICLR 2021's video + slides are here. Note this requires a login (registration).
M. Havasi, R. Jenatton, S. Fort, J. Z. Liu, J. Snoek, B. Lakshminarayanan, A. M. Dai, and D. Tran. Training independent subnetworks for robust prediction. In International Conference on Learning Representations, 2021.
@inproceedings{havasi2021training,
author = {Marton Havasi and Rodolphe Jenatton and Stanislav Fort and Jeremiah Zhe Liu and Jasper Snoek and Balaji Lakshminarayanan and Andrew M. Dai and Dustin Tran},
title = {Training independent subnetworks for robust prediction},
booktitle = {International Conference on Learning Representations},
year = {2021},
}