Machine Learning

Machine Learning is core to our products. Our researchers and engineers develop and deploy large scale ML and deep learning algorithms on one of the largest grid computing platforms in the world. We process over 100 billion daily user activities and over 200 billion daily ad requests over our global media and advertising networks.

Publications

Paper

June 6, 2024

Mathematical Programming Computation Journal

Machine Learning

Publications

Efficient Algorithms for Incremental Proximal Point Methods

Salient Object-Aware Background Generation Using Text-Guided Diffusion Models

Consistent Algorithms for Multi-Label Classification with Macro-at-K Metrics

SDEs for Minimax Optimization

Imposing Fairness Constraints in Synthetic Data Generation

Counterfactually Fair Representation

On Transfer of Adversarial Robustness from Pretraining to Downstream Tasks

Extreme Multi-Label Classification for Ad Targeting Using Factorization Machines

Loss Balancing for Fair Supervised Learning

Gradient Descent Converges Linearly for Logistic Regression on Separable Data

An SDE for Modeling SAM: Theory and Insights

Symbolic Metamodels for Interpreting Black-Boxes Using Primitive Functions

Precision/Recall on Imbalanced Test Data

Order-Invariant Cardinality Estimators Are Differentially Private

Incentive Mechanisms for Strategic Classification and Regression Problems

Fairness Interventions as (Dis)Incentives for Strategic Manipulation

On Missing Labels, Long-Tails, and Propensities in Extreme Multi-Label Classification

Set-Valued Prediction in Hierarchical Classification with Constrained Representation Complexity

Iterative Hard Thresholding with Adaptive Regularization: Sparser Solutions Without Sacrificing Runtime

Scalable Multi-Objective Optimization in Programmatic Advertising via Feedback Control

Hadoop-MTA: A System for Multi Data-Center Trillion Concepts Auto-ML Atop Hadoop

Transfer-Based Taxonomy Induction Over Concept Labels

Propensity-Scored Probabilistic Label Trees

Efficient Set-Valued Prediction in Multi-Class Classification

Online Probabilistic Label Trees

Local Search Algorithms for Rank-Constrained Convex Optimization

Effective Few Shot Classification with Transfer Learning

Locally Constructing Product Taxonomies from Scratch Using Representation Learning

Elucidation of DNA Methylation on N6-Adenine with Deep Learning

Sparse Convex Optimization via Adaptively Regularized Hard Thresholding

Edge Architecture for Dynamic Data Stream Analysis and Manipulation

Unsupervised Differentiable Multi-Aspect Network Embedding