2017
Report  Open Access

Exploring epoch-dependent stochastic residual networks

Carrara F., Esuli A., Falchi F., Moreo Fernández A.

Deep learning  Residual networks 

The recently proposed stochastic residual networks selectively activate or bypass the layers during training, based on independent stochastic choices, each of which following a probability distribution that is fixed in advance. In this paper we present a first exploration on the use of an epoch-dependent distribution, starting with a higher probability of bypassing deeper layers and then activating them more frequently as training progresses. Preliminary results are mixed, yet they show some potential of adding an epoch-dependent management of distributions, worth of further investigation.

Source: Research report, 2017



Back to previous page
BibTeX entry
@techreport{oai:it.cnr:prodotti:401323,
	title = {Exploring epoch-dependent stochastic residual networks},
	author = {Carrara F. and Esuli A. and Falchi F. and Moreo Fernández A.},
	institution = {Research report, 2017},
	year = {2017}
}
CNR ExploRA

Bibliographic record

ISTI Repository

Preprint version Open Access

Also available from

arxiv.orgOpen Access