search

terryum / awesome-deep-learning-papers

The most cited deep learning papers
25.36k stars 4.48k forks source link

Moving "Training Very Deep Networks" #57

Closed flukeskywalker closed 7 years ago

flukeskywalker commented 7 years ago

Our paper "Training Very Deep Networks" (NIPS 2015) is currently in the Appendix section since it has <200 citations. I propose that it should be moved to the main list.

Details: This paper is an extension of "Highway Networks" which appeared earlier at ICML DL workshop 2015. This is noted in the PDF: https://arxiv.org/pdf/1505.00387.pdf

Many authors prefer to cite the earliest paper, while others cite the NIPS version. There are some duplicates, but together these papers have over 300 citations as shown here:

Additionally, this paper has had wide impact: it is a predecessor of residual networks, demonstrates the relation between deep feedforward networks and LSTM networks, and is used in leading models for language modeling, NMT, speech recognition and generation etc. Therefore, I propose that it is reasonable to include this paper in the main list, as done in this PR.

flukeskywalker commented 7 years ago

@terryum It's been very long with no response on this. Any plans to resolve?