Deep neural networks include millions of learnable parameters, making their deployment over resource-constrained devices problematic. Sensitivity-based regularization of neurons (SeReNe) is a method for learning sparse topologies with a structure, exploiting neural sensitivity as a regularizer. We define the sensitivity of a neuron as the variation of the network output with respect to the variation of the activity of the neuron. The lower the sensitivity of a neuron, the less the network output is perturbed if the neuron output changes. By including the neuron sensitivity in the cost function as a regularization term, we are able to prune neurons with low sensitivity. As entire neurons are pruned rather than single parameters, practical network footprint reduction becomes possible. Our experimental results on multiple network architectures and datasets yield competitive compression ratios with respect to state-of-the-art references.

SeReNe: Sensitivity-Based Regularization of Neurons for Structured Sparsity in Neural Networks

Tartaglione, Enzo;Bragagnolo, Andrea;Odierna, Francesco;Fiandrotti, Attilio;Grangetto, Marco
2021

Abstract

Deep neural networks include millions of learnable parameters, making their deployment over resource-constrained devices problematic. Sensitivity-based regularization of neurons (SeReNe) is a method for learning sparse topologies with a structure, exploiting neural sensitivity as a regularizer. We define the sensitivity of a neuron as the variation of the network output with respect to the variation of the activity of the neuron. The lower the sensitivity of a neuron, the less the network output is perturbed if the neuron output changes. By including the neuron sensitivity in the cost function as a regularization term, we are able to prune neurons with low sensitivity. As entire neurons are pruned rather than single parameters, practical network footprint reduction becomes possible. Our experimental results on multiple network architectures and datasets yield competitive compression ratios with respect to state-of-the-art references.
1
14
https://ieeexplore.ieee.org/document/9456024
https://arxiv.org/pdf/2102.03773.pdf
Compression, deep networks, pruning, regularization, sparse networks.
Tartaglione, Enzo; Bragagnolo, Andrea; Odierna, Francesco; Fiandrotti, Attilio; Grangetto, Marco
File in questo prodotto:
File Dimensione Formato  
TNNLS21_SeReNe.pdf

accesso aperto

Tipo di file: POSTPRINT (VERSIONE FINALE DELL’AUTORE)
Dimensione 748 kB
Formato Adobe PDF
748 kB Adobe PDF Visualizza/Apri

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: http://hdl.handle.net/2318/1790716
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 3
  • ???jsp.display-item.citation.isi??? ND
social impact