Compression-aware Training of Deep Networks

Authors: Jose M. Alvarez, Mathieu Salzmann

NeurIPS 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments on Image Net and ICDAR show that we can achieve compression rates of more than 90%, thus hugely reducing the number of required operations at inference time.
Researcher Affiliation Collaboration Jose M. Alvarez Toyota Research Institute Los Altos, CA 94022 jose.alvarez@tri.global Mathieu Salzmann EPFL CVLab Lausanne, Switzerland mathieu.salzmann@epfl.ch
Pseudocode No The paper describes the approach using mathematical formulations and text, but does not include any pseudocode or algorithm blocks.
Open Source Code No The paper does not contain any explicit statement about releasing source code or a link to a code repository for the described methodology.
Open Datasets Yes For our experiments, we used two image classification datasets: Image Net [41] and ICDAR, the character recognition dataset introduced in [27].
Dataset Splits Yes We used the ILSVRC2012 [41] subset consisting of 1000 categories, with 1.2 million training images and 50,000 validation images.
Hardware Specification Yes For Image Net, training was done on a DGX-1 node using two-P100 GPUs in parallel. For ICDAR, we trained each network on a single Titan X-Pascal GPU for a total of 55 epochs...
Software Dependencies No More specifically, we used the torch-7 multi-gpu framework [11]. While a software name is provided, a specific version number for Torch7 is not mentioned in the text.
Experiment Setup Yes We used stochastic gradient descent with a momentum of 0.9 and a batch size of 180 images. The models were trained using an initial learning rate of 0.1 multiplied by 0.1 every 20 iterations... For ICDAR, we trained each network on a single Titan X-Pascal GPU for a total of 55 epochs with a batch size of 256 and 1,000 iterations per epoch.