Original science moves forward
The two studies explore a shared question at different scales: how to organize connectivity during learning. These scientific advances inform SparseMind’s work in models and systems engineering.
NeurIPS 2026
Two accepted papers.
One expanding research direction.
Two studies from the Cannistraci team have been accepted at NeurIPS 2026, advancing hardware-oriented semi-structured dynamic sparse training and channel-wise structural learning in convolutional networks.