back to list

Research Project: Sparse Neural Networks

Description

Training sparse artificial neural networks from scratch provides not only computational benefits contributing to greener AI, but often allows to achieve better generalisation than dense training. Since 2018 we witnessed the potential of SNN in a variety of learning scenarios and NN architectures, including but not limited to CNNs, LSTMs, GNNs, continual learning, and reinforcement learning. 

See also SNN Workshop series.

Details

Principal Investigator

Decebal Mocanu

Involved members:

Mykola Pechenizkiy

Bram Grooten

Tianjin Huang

Ghada Sokar

Qiao Xiao

Lu Yin