Picture for Amir Joudaki

Amir Joudaki

Emergence of Globally Attracting Fixed Points in Deep Neural Networks With Nonlinear Activations

Add code
Oct 29, 2024
Viaarxiv icon

Learning Genomic Sequence Representations using Graph Neural Networks over De Bruijn Graphs

Add code
Dec 06, 2023
Viaarxiv icon

Towards Training Without Depth Limits: Batch Normalization Without Gradient Explosion

Add code
Oct 03, 2023
Figure 1 for Towards Training Without Depth Limits: Batch Normalization Without Gradient Explosion
Figure 2 for Towards Training Without Depth Limits: Batch Normalization Without Gradient Explosion
Figure 3 for Towards Training Without Depth Limits: Batch Normalization Without Gradient Explosion
Figure 4 for Towards Training Without Depth Limits: Batch Normalization Without Gradient Explosion
Viaarxiv icon

On the impact of activation and normalization in obtaining isometric embeddings at initialization

Add code
May 28, 2023
Viaarxiv icon

Entropy Maximization with Depth: A Variational Principle for Random Neural Networks

Add code
May 25, 2022
Figure 1 for Entropy Maximization with Depth: A Variational Principle for Random Neural Networks
Viaarxiv icon

Batch Normalization Orthogonalizes Representations in Deep Random Networks

Add code
Jun 07, 2021
Figure 1 for Batch Normalization Orthogonalizes Representations in Deep Random Networks
Figure 2 for Batch Normalization Orthogonalizes Representations in Deep Random Networks
Figure 3 for Batch Normalization Orthogonalizes Representations in Deep Random Networks
Figure 4 for Batch Normalization Orthogonalizes Representations in Deep Random Networks
Viaarxiv icon