ちょっと情報が古いかつ一部異なるが、(主にサル用の)行動課題をこなすRNN(再帰的ニューラルネットワーク, Recurrent Neural Network)の論文集。
趣味研を始めるときや論文書き始めのときに調べました。
一部記憶はあやふやなのでご了承くださいm(. .)m
RNNが何をやっているかということへの仮説:課題の情報を高次元空間に振ってから低次元に圧縮してんじゃないか
Curr Opin Neurobiol. 2017 Oct;46:1-6. doi: 10.1016/j.conb.2017.06.003. Epub 2017 Jun 29.
Recurrent neural networks as versatile tools of neuroscience research.
Barak O1.
https://www.ncbi.nlm.nih.gov/pubmed/28668365
サルの課題をRNNに学習させるとサルのPFCの挙動と似た挙動を示した
Nature. 2013 Nov 7;503(7474):78-84. doi: 10.1038/nature12742.
Context-dependent computation by recurrent dynamics in prefrontal cortex.
Mante V1, Sussillo D, Shenoy KV, Newsome WT.
https://www.ncbi.nlm.nih.gov/pubmed/24201281
RNN内の神経素子を興奮性と抑制性に分けても学習できた
PLoS Comput Biol. 2016 Feb 29;12(2):e1004792. doi: 10.1371/journal.pcbi.1004792. eCollection 2016.
Training Excitatory-Inhibitory Recurrent Neural Networks for Cognitive Tasks: A Simple and Flexible Framework.
Song HF1, Yang GR1, Wang XJ1,2.
https://www.ncbi.nlm.nih.gov/pubmed/26928718
https://github.com/xjwanglab/pycog/tree/master/pycog
強化学習ベースのRNNシステムで行動課題を学習
BioRxiv-> eLife
Reward-based training of recurrent neural networks for cognitive and value-based tasks
H. Francis Song, Guangyu R. Yang, Xiao-Jing Wang
http://biorxiv.org/content/early/2016/08/19/070375
http://github.com/xjwanglab/pyrl
https://elifesciences.org/content/6/e21492
Hebb則(亜種)+強化学習のRNNシステムで行動課題を学習
BioRxiv -> eLife
Flexible decision-making in recurrent neural networks trained with a biologically plausible rule
Thomas Miconi
http://biorxiv.org/content/early/2016/07/26/057729
https://github.com/ThomasMiconi/BiologicallyPlausibleLearningRNN
https://elifesciences.org/content/6/e20899
連続的な行動をRNNで学習できた
Neuron. 2016 Apr 6;90(1):128-42. doi: 10.1016/j.neuron.2016.02.009. Epub 2016 Mar 10.
Recurrent Network Models of Sequence Generation and Memory.
Rajan K1, Harvey CD2, Tank DW3.
https://www.ncbi.nlm.nih.gov/pubmed/26971945
課題への回答の確信度みたいのも学習できた(OFCと挙動が似てた?*要見直し)
Neuron. 2015 May 20;86(4):1067-77. doi: 10.1016/j.neuron.2015.04.014. Epub 2015 May 7.
Dynamic Control of Response Criterion in Premotor Cortex during Perceptual Detection under Temporal Uncertainty.
Carnevale F1, de Lafuente V2, Romo R3, Barak O4, Parga N1.
https://www.ncbi.nlm.nih.gov/pubmed/25959731
RNNでMEG(筋電位)の挙動を再現
Nat Neurosci. 2015 Jul;18(7):1025-33. doi: 10.1038/nn.4042. Epub 2015 Jun 15.
A neural network that finds a naturalistic solution for the production of muscle activity.
Sussillo D1, Churchland MM2, Kaufman MT1, Shenoy KV3.
https://www.ncbi.nlm.nih.gov/pubmed/26075643
RNNで行動課題ができた(再帰結合はへんかしない?*要見直し)
J Neurosci. 2013 Jul 10;33(28):11515-29. doi: 10.1523/JNEUROSCI.5044-12.2013.
Emergence of dynamic memory traces in cortical microcircuit models through STDP
Klampfl S1, Maass W.
https://www.ncbi.nlm.nih.gov/pubmed/23843522
Cereb Cortex. 2014 Mar;24(3):677-90. doi: 10.1093/cercor/bhs348. Epub 2012 Nov 11.
Emergence of complex computational structures from chaotic neural networks through reward-modulated Hebbian learning.
Hoerzer GM1, Legenstein R, Maass W.
https://www.ncbi.nlm.nih.gov/pubmed/23146969
Continual learning(ある課題を学習させた後に別の課題を学習させても、前の課題の情報を失わないようにさせた学習方法)なRNNに複数の課題を学習させると課題間の関係がベクトル表現できるようになった
BioRxiv
Clustering and compositionality of task representations in a neural network trained to perform many cognitive tasks
Guangyu Robert Yang, H. Francis Song, William T. Newsome, Xiao-Jing Wang
https://www.biorxiv.org/content/early/2017/09/01/183632