Learning to Compose Neural Networks for Question Answering · HackerPapers discussions

Learning to Compose Neural Networks for Question Answering — mentioned by KhoomeiK

Tue, 08 Dec 2020 21:24:03 GMT

Discovering Symbolic Models from Deep Learning with Inductive Biases [1] trains graph neural nets on astrophysical phenomena and then performs symbolic regression to generate algebraic formulae to elegantly model the phenomena in a classical physics framework. It's largely gone under the radar but has pretty interesting implications for NLP and language theory in my opinion. Direct Feedback Alignment Scales to Modern Deep Learning Tasks and Architectures [2] applies DFA, an approach to training neural nets without backprop, to modern architectures like the Transformer. It does surprisingly…

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

Learning to Compose Neural Networks for Question Answering — mentioned by daddyo

Mon, 10 Jul 2017 23:00:30 GMT

Yes. I think these architectures are very exciting and a step in the "right" direction. Eventually we will want to move from rote memorization and pattern matching to more challenging aspects of intelligence. https://arxiv.org/abs/1601.01705v4 (Learning to Compose Neural Networks for Question Answering) comes close to breaking this barrier.

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

Learning to Compose Neural Networks for Question Answering — HN submission

Mon, 02 May 2016 15:32:08 GMT

Learning to Compose Neural Networks for Question Answering

3 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.