Deploying large convolutional neural networks (CNNs) on limited-resource devices is still an open challenge in the big data era. To deal with this challenge, a synergistic composition of network compression algorithms and compact storage of the compressed network has been recently presented, substantially preserving model accuracy. The proposed implementation, which we describe in this paper, offers different compression schemes (pruning, two types of weight quantization, and their combinations) and two compact representations: the Huffman Address Map compression (HAM), and its sparse version sHAM. Taken as input a model, trained for a given classification or regression problem (as well as the dataset employed, which is necessary for the fine-tuning of weights after network compression), the procedure returns the corresponding compressed model. Our publicly available implementation provides the source code, two pre-trained CNN models (retrieved from third-party repositories referring to well-established literature), and four datasets. This implementation includes detailed instructions to execute the scripts and reproduce the obtained results, in terms of the figures and tables included in the original paper.

Reproducing the Sparse Huffman Address Map Compression for Deep Neural Networks / G.C. Marino, G. Ghidoli, M. Frasca, D. Malchiodi (LECTURE NOTES IN ARTIFICIAL INTELLIGENCE). - In: Reproducible Research in Pattern Recognition / [a cura di] B. Kerautret, M. Colom, A. Krähenbühl, D. Lopresti, P. Monasse, H. Talbot. - [s.l] : Springer, 2021. - ISBN 978-3-030-76422-7. - pp. 161-166 (( Intervento presentato al 3. convegno International Workshop on Reproducible Research in Pattern Recognition, RRPR 2021 tenutosi a Milano nel 2021 [10.1007/978-3-030-76423-4_12].

Reproducing the Sparse Huffman Address Map Compression for Deep Neural Networks

M. Frasca;D. Malchiodi
2021

Abstract

Deploying large convolutional neural networks (CNNs) on limited-resource devices is still an open challenge in the big data era. To deal with this challenge, a synergistic composition of network compression algorithms and compact storage of the compressed network has been recently presented, substantially preserving model accuracy. The proposed implementation, which we describe in this paper, offers different compression schemes (pruning, two types of weight quantization, and their combinations) and two compact representations: the Huffman Address Map compression (HAM), and its sparse version sHAM. Taken as input a model, trained for a given classification or regression problem (as well as the dataset employed, which is necessary for the fine-tuning of weights after network compression), the procedure returns the corresponding compressed model. Our publicly available implementation provides the source code, two pre-trained CNN models (retrieved from third-party repositories referring to well-established literature), and four datasets. This implementation includes detailed instructions to execute the scripts and reproduce the obtained results, in terms of the figures and tables included in the original paper.
CNN compression; Weight pruning; Weight sharing; Probabilistic quantization; Entropy coding
Settore INF/01 - Informatica
Book Part (author)
File in questo prodotto:
File Dimensione Formato  
published-RRPR21.pdf

accesso riservato

Tipologia: Publisher's version/PDF
Dimensione 1.06 MB
Formato Adobe PDF
1.06 MB Adobe PDF   Visualizza/Apri   Richiedi una copia
Pubblicazioni consigliate

Caricamento pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/2434/876040
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? ND
social impact