Implicit Bias

Implicit bias refers to unintended, often subtle, biases embedded within machine learning models, stemming from biases present in their training data. Current research focuses on detecting and mitigating these biases in various model architectures, particularly large language models (LLMs) and deep neural networks, using techniques like prompt engineering, fine-tuning, and Bayesian methods. Understanding and addressing implicit bias is crucial for ensuring fairness and equity in AI applications, impacting fields ranging from healthcare and criminal justice to education and hiring. The development of robust bias detection and mitigation strategies is a central goal of ongoing research.

Papers

June 30, 2023

The Implicit Bias of Minima Stability in Multivariate Shallow ReLU Networks
Mor Shpigel Nacson, Rotem Mulayoff, Greg Ongie, Tomer Michaeli, Daniel Soudry
Implicit Bias ReLU Network Shallow Network Flat Minimum Shallow ReLU Layer ReLU Network

June 20, 2023

The Implicit Bias of Batch Normalization in Linear Models and Two-layer Linear Convolutional Neural Networks
Yuan Cao, Difan Zou, Yuanzhi Li, Quanquan Gu
Stochastic Gradient Descent Batch Normalization Implicit Bias Deep Linear Linear Model Margin Classifier Linear Neural Network

June 14, 2023

Beyond Implicit Bias: The Insignificance of SGD Noise in Online Learning
Nikhil Vyas, Depen Morwani, Rosie Zhao, Gal Kaplun, Sham Kakade, Boaz Barak
Deep Learning Stochastic Gradient Descent Online Learning Implicit Bias Offline Learning Gradient Step Rating Meaningfulness Multi Epoch Training

June 10, 2023

Learning a Neuron by a Shallow ReLU Network: Dynamics and Implicit Bias for Correlated Inputs
Dmitry Chistikov, Matthias Englert, Ranko Lazic
LeArning Abstract Gradient Flow Single Neuron Level Implicit Bias Teacher Network Shallow ReLU Single Neuron Layer ReLU Network Neural Interpolation

June 1, 2023

Combining Explicit and Implicit Regularization for Efficient Learning in Deep Networks
Dan Zhao
Deep Network Efficient Learning Implicit Bias Implicit Regularization Deep Linear Explicit Regularization Explicit Point Solo

May 25, 2023

Implicit bias of SGD in $L_{2}$-regularized linear DNNs: One-way jumps from high to low rank
Zihan Wang, Arthur Jacot
Stochastic Gradient Descent Low Rank Implicit Bias Matrix Completion Hidden Layer Deep Linear Linear Neural Network Local Minimum

May 24, 2023

May 19, 2023

Implicit Bias of Gradient Descent for Logistic Regression at the Edge of Stability
Jingfeng Wu, Vladimir Braverman, Jason D. Lee
Gradient Descent Extreme Edge Core Stability Implicit Bias Logistic Regression Logistics System Unbounded Loss

May 16, 2023

Measuring Implicit Bias Using SHAP Feature Importance and Fuzzy Cognitive Maps
Isel Grau, Gonzalo Nápoles, Fabian Hoitsma, Lisa Koutsoviti Koumeri, Koen Vanhoof
Feature Importance Implicit Bias Neural Process Fuzzy Cognitive Map

March 14, 2023

On the Implicit Geometry of Cross-Entropy Parameterizations for Label-Imbalanced Data
Tina Behnia, Ganesh Ramachandra Kini, Vala Vakilian, Christos Thrampoulidis
Implicit Bias Cross Entropy Logit Adjustment

March 7, 2023

On the Implicit Bias of Linear Equivariant Steerable Networks
Ziyu Chen, Wei Zhu
Gradient Flow Implicit Bias Steerable Convolution Group Invariant

March 2, 2023

The Double-Edged Sword of Implicit Bias: Generalization vs. Robustness in ReLU Networks
Spencer Frei, Gal Vardi, Peter L. Bartlett, Nathan Srebro
Native Robustness Strong Generalization Adversarial Example Adversarial Robustness Gradient Flow Implicit Bias ReLU Network Two Layer ReLU

January 30, 2023

On student-teacher deviations in distillation: does it pay to disobey?
Vaishnavh Nagarajan, Aditya Krishna Menon, Srinadh Bhojanapalli, Hossein Mobahi, Sanjiv Kumar
Knowledge Distillation Implicit Bias Better Generalization Low Temperature Distillation Improved Generalization Teacher Student

December 28, 2022

On Implicit Bias in Overparameterized Bilevel Optimization
Paul Vicol, Jonathan Lorraine, Fabian Pedregosa, David Duvenaud, Roger Grosse
Bilevel Optimization Implicit Bias Gradient Based Algorithm Single Level Optimization

December 5, 2022

Can Ensembling Pre-processing Algorithms Lead to Better Machine Learning Fairness?
Khaled Badran, Pierre-Olivier Côté, Amanda Kolopanis, Rached Bouchoucha, Antonio Collante, Diego Elias Costa, Emad Shihab, Foutse Khomh
Implicit Bias Model Ensembling Fair Algorithm Preprocessing Method Machine Learning Fairness

October 25, 2022

Same Pre-training Loss, Better Downstream: Implicit Bias Matters for Language Models
Hong Liu, Sang Michael Xie, Zhiyuan Li, Tengyu Ma
Language Model Pre Training Implicit Bias Autoregressive Language Model Interest Loss Downstream Performance

October 13, 2022

Implicit Bias in Leaky ReLU Networks Trained on High-Dimensional Data
Spencer Frei, Gal Vardi, Peter L. Bartlett, Nathan Srebro, Wei Hu
Gradient Descent Low Rank Gradient Flow High Dimensional Data Implicit Bias Fully Connected Neural Network Homogeneous Neural Network Leaky ReLU

October 6, 2022

Understanding Gradient Regularization in Deep Learning: Efficient Finite-Difference Computation and Implicit Bias
Ryo Karakida, Tomoumi Takase, Tomohiro Hayase, Kazuki Osawa
Deep Learning Gradient Norm Implicit Bias Gradient Regularization Finite Difference Gradient Ascent

September 30, 2022

Self-Stabilization: The Implicit Bias of Gradient Descent at the Edge of Stability
Alex Damian, Eshaan Nichani, Jason D. Lee
Gradient Descent Extreme Edge Core Stability Implicit Bias Large Batch Automatic Stabilization

Implicit Bias

Papers

The Implicit Bias of Minima Stability in Multivariate Shallow ReLU Networks

The Implicit Bias of Batch Normalization in Linear Models and Two-layer Linear Convolutional Neural Networks

Beyond Implicit Bias: The Insignificance of SGD Noise in Online Learning

Learning a Neuron by a Shallow ReLU Network: Dynamics and Implicit Bias for Correlated Inputs

Combining Explicit and Implicit Regularization for Efficient Learning in Deep Networks

Implicit bias of SGD in $L_{2}$-regularized linear DNNs: One-way jumps from high to low rank

Relating Implicit Bias and Adversarial Attacks through Intrinsic Dimension

Getting Sick After Seeing a Doctor? Diagnosing and Mitigating Knowledge Conflicts in Event Temporal Reasoning

Implicit Bias of Gradient Descent for Logistic Regression at the Edge of Stability

Measuring Implicit Bias Using SHAP Feature Importance and Fuzzy Cognitive Maps

On the Implicit Geometry of Cross-Entropy Parameterizations for Label-Imbalanced Data

On the Implicit Bias of Linear Equivariant Steerable Networks

The Double-Edged Sword of Implicit Bias: Generalization vs. Robustness in ReLU Networks

On student-teacher deviations in distillation: does it pay to disobey?

On Implicit Bias in Overparameterized Bilevel Optimization

Can Ensembling Pre-processing Algorithms Lead to Better Machine Learning Fairness?

Same Pre-training Loss, Better Downstream: Implicit Bias Matters for Language Models

Implicit Bias in Leaky ReLU Networks Trained on High-Dimensional Data

Understanding Gradient Regularization in Deep Learning: Efficient Finite-Difference Computation and Implicit Bias

Self-Stabilization: The Implicit Bias of Gradient Descent at the Edge of Stability