Login / Signup
JNPU: A 1.04TFLOPS Joint-DNN Training Processor with Speculative Cyclic Quantization and Triple Heterogeneity on Microarchitecture / Precision / Dataflow.
Je Yang
Sukbin Lim
Sukjin Lee
Jae-Young Kim
Joo-Young Kim
Published in:
ESSCIRC (2023)
Keyphrases
</>
training process
parallel processing
training set
high speed
precision and recall
computer architecture
supervised learning
training samples
training examples
training phase
training algorithm
high precision
motion estimation
multiscale
early stage
data sets
image quality
parallel computing
support vector machine