Low-memory convolutional neural networks through incremental depth-first processing

04/28/2018
by   Jonathan Binas, et al.
0

We introduce an incremental processing scheme for convolutional neural network (CNN) inference, targeted at embedded applications with limited memory budgets. Instead of processing layers one by one, individual input pixels are propagated through all parts of the network they can influence under the given structural constraints. This depth-first updating scheme comes with hard bounds on the memory footprint: the memory required is constant in the case of 1D input and proportional to the square root of the input dimension in the case of 2D input.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
04/16/2018

Training convolutional neural networks with megapixel images

To train deep convolutional neural networks, the input data and the inte...
research
05/09/2017

Model Complexity-Accuracy Trade-off for a Convolutional Neural Network

Convolutional Neural Networks(CNN) has had a great success in the recent...
research
05/28/2018

Convolutional neural network compression for natural language processing

Convolutional neural networks are modern models that are very efficient ...
research
08/12/2021

perf4sight: A toolflow to model CNN training performance on Edge GPUs

The increased memory and processing capabilities of today's edge devices...
research
08/08/2017

Prune the Convolutional Neural Networks with Sparse Shrink

Nowadays, it is still difficult to adapt Convolutional Neural Network (C...
research
05/03/2018

How deep should be the depth of convolutional neural networks: a backyard dog case study

We present a straightforward non-iterative method for shallowing of deep...
research
09/06/2017

Embedded Binarized Neural Networks

We study embedded Binarized Neural Networks (eBNNs) with the aim of allo...

Please sign up or login with your details

Forgot password? Click here to reset