Improved Projection Learning for Lower Dimensional Feature Maps

10/27/2022
by   Ilan Price, et al.
0

The requirement to repeatedly move large feature maps off- and on-chip during inference with convolutional neural networks (CNNs) imposes high costs in terms of both energy and time. In this work we explore an improved method for compressing all feature maps of pre-trained CNNs to below a specified limit. This is done by means of learned projections trained via end-to-end finetuning, which can then be folded and fused into the pre-trained network. We also introduce a new `ceiling compression' framework in which evaluate such techniques in view of the future goal of performing inference fully on-chip.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
10/12/2021

Memory-Efficient CNN Accelerator Based on Interlayer Feature Map Compression

Existing deep convolutional neural networks (CNNs) generate massive inte...
research
07/23/2020

End-to-end Learning of Compressible Features

Pre-trained convolutional neural networks (CNNs) are powerful off-the-sh...
research
09/25/2019

CAT: Compression-Aware Training for bandwidth reduction

Convolutional neural networks (CNNs) have become the dominant neural net...
research
10/31/2017

Clothing Retrieval with Visual Attention Model

Clothing retrieval is a challenging problem in computer vision. With the...
research
05/30/2018

On the Spectrum of Random Features Maps of High Dimensional Data

Random feature maps are ubiquitous in modern statistical machine learnin...
research
08/08/2021

Combining machine learning and data assimilation to forecast dynamical systems from noisy partial observations

We present a supervised learning method to learn the propagator map of a...
research
04/18/2021

Barrier-Free Large-Scale Sparse Tensor Accelerator (BARISTA) For Convolutional Neural Networks

Convolutional neural networks (CNNs) are emerging as powerful tools for ...

Please sign up or login with your details

Forgot password? Click here to reset