Hemanth Saratchandran

9

Papers

38

Total Citations

Papers (9)

Efficient Learning with Sine-Activated Low-Rank Matrices

From Activation to Initialization: Scaling Insights for Optimizing Neural Fields

Always Skip Attention

Towards Higher Effective Rank in Parameter-Efficient Fine-tuning using Khatri-Rao Product

Preconditioners for the Stochastic Training of Neural Fields

Enhancing Transformers Through Conditioned Embedded Tokens

A sampling theory perspective on activations for implicit neural representations

Curvature-Aware Training for Coordinate Networks

On skip connections and normalisation layers in deep optimisation