Picture for Elvis Dohmatob

Elvis Dohmatob

PARIETAL

auto-fpt: Automating Free Probability Theory Calculations for Machine Learning Theory

Add code
Apr 14, 2025
Viaarxiv icon

Improving the Scaling Laws of Synthetic Data with Deliberate Practice

Add code
Feb 21, 2025
Viaarxiv icon

The Pitfalls of Memorization: When Memorization Hurts Generalization

Add code
Dec 10, 2024
Viaarxiv icon

Strong Model Collapse

Add code
Oct 07, 2024
Figure 1 for Strong Model Collapse
Figure 2 for Strong Model Collapse
Figure 3 for Strong Model Collapse
Figure 4 for Strong Model Collapse
Viaarxiv icon

Beyond Model Collapse: Scaling Up with Synthesized Data Requires Reinforcement

Add code
Jun 11, 2024
Figure 1 for Beyond Model Collapse: Scaling Up with Synthesized Data Requires Reinforcement
Figure 2 for Beyond Model Collapse: Scaling Up with Synthesized Data Requires Reinforcement
Figure 3 for Beyond Model Collapse: Scaling Up with Synthesized Data Requires Reinforcement
Figure 4 for Beyond Model Collapse: Scaling Up with Synthesized Data Requires Reinforcement
Viaarxiv icon

Model Collapse Demystified: The Case of Regression

Add code
Feb 12, 2024
Viaarxiv icon

A Tale of Tails: Model Collapse as a Change of Scaling Laws

Add code
Feb 10, 2024
Figure 1 for A Tale of Tails: Model Collapse as a Change of Scaling Laws
Figure 2 for A Tale of Tails: Model Collapse as a Change of Scaling Laws
Figure 3 for A Tale of Tails: Model Collapse as a Change of Scaling Laws
Figure 4 for A Tale of Tails: Model Collapse as a Change of Scaling Laws
Viaarxiv icon

Scaling Laws for Associative Memories

Add code
Oct 04, 2023
Viaarxiv icon

Robust Linear Regression: Phase-Transitions and Precise Tradeoffs for General Norms

Add code
Aug 01, 2023
Viaarxiv icon

Robust Linear Regression: Gradient-descent, Early-stopping, and Beyond

Add code
Jan 31, 2023
Figure 1 for Robust Linear Regression: Gradient-descent, Early-stopping, and Beyond
Figure 2 for Robust Linear Regression: Gradient-descent, Early-stopping, and Beyond
Figure 3 for Robust Linear Regression: Gradient-descent, Early-stopping, and Beyond
Figure 4 for Robust Linear Regression: Gradient-descent, Early-stopping, and Beyond
Viaarxiv icon