Login / Signup
A 2Mbit Digital in-Memory Computing Matrix-Vector Multiplier for DNN Inference supporting flexible bit precision and matrix size achieving 612 binary TOPS/W.
Mohit Gupta
Stefan Cosemans
Peter Debacker
Wim Dehaene
Published in:
ESSCIRC (2023)
Keyphrases
</>
sparse matrix
matrix representation
covariance matrix
memory requirements
memory size
memory usage
training data
feature space
precision and recall
memory space
singular values
inference process
transformation matrix