Krzysztof Marcin Choromanski

Graph Random Features for Scalable Gaussian Processes

Computationally Efficient Graph Modelling with Refined Graph Random Features

Wavelet-Induced Rotary Encodings: RoPE Meets Graphs

Linear Transformer Topological Masking with Graph Random Features

Towards Scalable Exact Machine Unlearning Using Parameter-Efficient Fine-Tuning

Learning the RoPEs: Better 2D and 3D Position Encodings with STRING

Variance-Reducing Couplings for Random Features

EUGens: Efficient, Unified and General Dense Layers

Magnituder Layers for Implicit Neural Representations in 3D