Problem sheet

IID Client Partitioning

federated-learning · partitioning · iid

Weighted Federated Averaging

fedavg · aggregation · weighted-average

Client Updates as Deltas

fedavg · model-updates · server-optimizer

Label-Skewed (non-IID) Partition

non-iid · label-skew · heterogeneity

Client Training with a PyTorch Model

pytorch · federated-learning · local-training

Local SGD on a ClientOptional

local-training · sgd · minibatch

One FedAvg Round

fedavg · communication-round · federated-learning

Seeded Client Sampling

client-sampling · participation · cross-device

Clipping Client Updates

clipping · robustness · privacy

Full FedAvg Training Loop

fedavg · training-loop · convergence

FedAvg with PyTorch state_dicts

pytorch · federated-learning · fedavg

Secure Aggregation with Pairwise Masks

secure-aggregation · privacy · masking

Differentially Private Aggregation

differential-privacy · gaussian-mechanism · clipping

FedProx Local Objective

fedprox · non-iid · client-drift

Federated Training across Departments

Project idea · federated-learning · non-iid · flower