EPFL Machine Learning and Optimization Laboratory

sent2vec

General purpose unsupervised sentence representations

ChocoSGD

Decentralized SGD and Consensus with Communication Compression: https://arxiv.org/abs/1907.09356

sparsifiedSGD

Sparsified SGD with Memory: https://arxiv.org/abs/1809.07599

LocalSGD-Code

byzantine-robust-optimizer

Learning from history for Byzantine Robustness

interpret-lm-knowledge

Extracting knowledge graphs from language models as a diagnostic benchmark of model performance (NeurIPS XAI 2021).

Bi-Sent2Vec

Robust Cross-lingual Embeddings from Parallel Sentences

opt-summerschool

Short Course on Optimization for Machine Learning - Slides and Practical Labs - DS3 Data Science Summer School, June 24 to 28, 2019, Paris, France

cola

CoLa - Decentralized Linear Learning: https://arxiv.org/abs/1808.04883

opt-shortcourse

Short Course on Optimization for Machine Learning - Slides and Practical Lab - Pre-doc Summer School on Learning Systems, July 3 to 7, 2017, Zürich, Switzerland

byzantine-robust-noniid-optimizer

powergossip

Code for "Practical Low-Rank Communication Compression in Decentralized Deep Learning"

X2Static

X2Static embeddings

kubernetes-setup

MLO group setup for kubernetes cluster

topology-in-decentralized-learning

Code related to ’Beyond spectral gap: The role of the topology in decentralized learning‘.

byzantine-robust-decentralized-optimizer

uncertainity-estimation

Code for the paper “The Peril of Popular Deep Learning Uncertainty Estimation Methods”

FeAI

Federated Learning with TensorFlow.js

autoTrain

Open Challenge - Automatic Training for Deep Learning

pax

JAX-like API for PyTorch

ContinuityPlan

Continuity Plan in accordance with https://www.epfl.ch/campus/security-safety/en/health/coronavirus-covid19/

phantomedicus

MedSurge: medical survey generator

GradientAlignment

correlating-tweets

Correlating Twitter Language with Community-Level Health Outcomes: https://arxiv.org/abs/1906.06465

personalized-collaborative-speedup

code for the paper "Linear Speedup in Personalized Collaborative Learning"

SemEval2016

Code for SemEval-2016 winning classifier "SwissCheese at SemEval-2016 Task 4: Sentiment Classification Using an Ensemble of Convolutional Neural Networks with Distant Supervision"

tensorflow

extension of distributed training to sparse linear models (L1 regularizers, using primal CoCoA, instead of dual CoCoA for the L2 case in default tensorflow), by @LiamHe