Random feedback weights support learning in deep neural networks · HackerPapers discussions

Random feedback weights support learning in deep neural networks — mentioned by etiam

Tue, 21 Mar 2017 18:22:23 GMT

The Hinton reference you were trying to find was likely joint work with Timothy Lillicrap. Geoff refers to it, among other places, in the (recorded) lecure from the reminescence symposium for David MacKay. http://divf.eng.cam.ac.uk/djcms2016/#hinton I don't know that there is an article out for that work yet, but the initial finding by Lilllicrap et al. has been published as a preprint: https://arxiv.org/abs/1411.0247

Paper and discussion history. A mention is not an endorsement.

Random feedback weights support learning in deep neural networks — mentioned by taliesinb

Tue, 30 Aug 2016 08:45:06 GMT

Are you referring to http://arxiv.org/pdf/1411.0247v1.pdf ? That doesn't mention sparse backward connections, but it does show that feedback weights that aren't computing the actual derivative dloss/din can till support learning. The network 'learns to learn', so to speak.

Paper and discussion history. A mention is not an endorsement.

Random feedback weights support learning in deep neural networks — mentioned by p1esk

Fri, 25 Dec 2015 01:36:13 GMT

No, we don't know that. Take a look at this paper: http://arxiv.org/abs/1411.0247

Paper and discussion history. A mention is not an endorsement.

Random feedback weights support learning in deep neural networks — HN submission

Fri, 28 Nov 2014 14:43:20 GMT

Random feedback weights support learning in deep neural networks

77 points; 14 discussion replies on HN. Paper and source history. A submission is not an endorsement.