Search papers, labs, and topics across Lattice.
1
0
3
Quantized neural nets can be made 15% more accurate in the face of bit errors, just by swapping in a new loss function.
