Search papers, labs, and topics across Lattice.
Aarhus University
2
0
3
Sparse, quantized linear-attention models can outperform dense counterparts on neuromorphic hardware, achieving up to 37脳 higher throughput and 16脳 lower power consumption.
Allocating synaptic weights based on neuron firing rates can slash power consumption by over 60% in Spiking Neural Networks without significant area costs.
