Strong Lottery Ticket Hypothesis

Algorithm and hardware co-design for subnetworks of Strong Lottery Ticket Hypothesis in FeFET-based Computation-in-Memory at edge.

Strong Lottery Ticket Hypothesis (SLTH) algorithm is a neural network with initial random weights and has subnetworks that get the correct answer. SLTH requires a mechanism for using or not using weights (connecting or not connecting synapses). Ferroelectric FET-based Computation-in-Memory (FeFET CiM) uses FeFET cells to store weights and the other FeFET cells to transfer weight information. This achieves a co-design that makes the SLTH algorithm computable at the edge.