pySigLib - Fast Signature-Based Computations on CPU and GPU
Daniil Shmelev ⋅ Cristopher Salvi
Abstract
Signature-based methods have recently gained significant traction in machine learning for sequential data. However, existing implementations do not scale to the dataset sizes and sequence lengths encountered in practice. We present pySigLib, a high-performance toolkit for rough path computations on CPUs and GPUs. pySigLib offers native integration with PyTorch and JAX, enabling the adoption of signature-based methods within standard machine learning pipelines.
Chat is not available.
Successful Page Load