Unbiased federated learning model for skin lesion classification

Federated Learning

Multimodal images

Domain Adaptation

Debiased Method

TABE: use custom loss which minimize the main classification loss, and train additional classifier using confusion loss between additional class label and uniform distribution with the same feature representation

LNTL: use adversarial and gradient reversal method to train two classifier, first one is the main classifier, the second one is domain classifier with gradient reversal

Detecting Melanoma Fairly: Skin Tone Detection and Debiasing for Skin Lesion Classification

Additional fusion branch

Multi-scale Fully-shared Fusion Network Multi-scale fusion structure combines deep and shallow features within individual modalities, and Dermo-Clinical Block (DCB) integrates the feature maps from dermoscopic images and clinical images

Multi-Label classification of multi-modality skin lesion via hyper-connected convolutional neural network

Domain-Adversarial Training of Neural Networks (Gradient Reversal) Train a domain classifier and multiplies the gradient by a certain negative constant during the backpropagation-based training on feature extractor layer's weights.

Late fusion

Seven-Point Checklist and Skin Lesion Classification Using Multitask Multimodal Neural Nets

Multimodal skin lesion classification using deep learning

Co-Attention Fusion Network for Multimodal Skin Cancer Diagnosis

Performance Measurement

Accuracy benefit of FL: accuracy of (FL model - local model) on local test set

Fairness benefit of FL: fairness of (FL model - local model) on local test set. Fairness is measured using equalized odds (Hardt et al., 2016) and demographic parity (Dwork et al., 2012)

Communication efficiency

Local updating

FedAvg (McMahan 2017)

FedProx (Li 2018)

Compression schemes

Structured & sketched updates (Konecny 2016)

Lossy compression & dropout (Caldas 2018)

Sparse Ternary Compression (Sattler 2019)

Decentralized training

Systems heterogeneity

Decentralized Linear Learning (He 2018)

Topology

FeSTA: Federated split learning (Park 2021)

FesViBS: Festa with block sampling (Almalik 2023)

Statistical heterogeneity

Asynchronous communication

Active sampling

Device sampling based on systems resources (Nishio 2018)

Incentive design using contract theory (Kang 2019)

Fault tolerance

Ignoring the stragglers (Chen 2016)

Allow for low participation FedProx (Li 2018)

Replicating data blocks and coding across gradients (Tandon 2017)

Modeling heterogeneous data

MOCHA: Federated Multi-Task Learning (Smith 2017)

Star topology using Variational Federated Multi-Task Learning (Corinzia 2019)

ARUBA: Adaptive Gradient-Based Meta-Learning Methods (Khodak 2019)

Semi-Cyclic Stochastic Gradient Descent: a pluraistic solution (Eichner 2019)

Weights initialization on shared proxy data (Zhao 2018)

Bias reduction

Reveal bias

(De)Constructing Bias on Skin Lesion Datasets (Bissoto 2019)

LoAdaBoost: Perform more local training based on loss (Huang 2018)

Agnostic Federated Learning (Mohri 2019)

Fairness

Uniformity in performance of model

q-FFL: Higher weight for higher loss clients (Li 2020)

Refine client update

SCAFFOLD: Stochastic Controlled Averaging for Federated Learning (Karimireddy 2019)

Regularize the local objective function

FedProx (Li 2018)

FedDyn: Federated Learning Based on Dynamic Regularization (Durmus 2021)

Personalized FL

click to edit