A library built upon PyTorch for multimodal learning and transfer learning from multiple data sources
Make abundant machine learning software accessible for interdisciplinary research
Reduce repetitions, reuse resources, and recycle models to build PyKale
PyKale is a Python library providing accessible machine learning from multiple data sources for interdisciplinary research, particularly multimodal learning and transfer learning, named collectively as Knowledge-aware machine learning (Kale).
β Listed in the official PyTorch Landscape as one of only four libraries under “Multimodal”, alongside MMF (Meta), NeMo (NVIDIA), and USB (Microsoft).
Learn from data of multiple sources (modalities / domains) under one roof.
Separate code and configurations for non-programmers to configure systems without coding.
All machine learning workflows follow a standardized six-step pipeline.
Developed mainly at the University of Sheffield, with partial support from an EPSRC NetworkPlus grant (UKRI396), a Wellcome Trust Innovator Award (215799/Z/19/Z), and university funding through the Centre for Machine Intelligence and its AI Research Engineering (AIRE) team.
