Matches in SemOpenAlex for { <https://semopenalex.org/work/W4367172263> ?p ?o ?g. }
- W4367172263 abstract "Abstract A key feature of animal and human decision-making is to balance the exploration of unknown options for information gain (directed exploration) versus selecting known options for immediate reward (exploitation), which is often examined using restless bandit tasks. Recurrent neural network models (RNNs) have recently gained traction in both human and systems neuroscience work on reinforcement learning, due to their ability to show meta-learning of task domains. Here we comprehensively compared the performance of a range of RNN architectures as well as human learners on restless four-armed bandit problems. The best-performing architecture (LSTM network with computation noise) exhibited human-level performance. Cognitive modeling first revealed that both human and RNN behavioral data contain signatures of higher-order perseveration, i.e., perseveration beyond the last trial, but this effect was more pronounced in RNNs. In contrast, human learners, but not RNNs, exhibited a positive effect of uncertainty on choice probability (directed exploration). RNN hidden unit dynamics revealed that exploratory choices were associated with a disruption of choice predictive signals during states of low state value, resembling a win-stay-loose-shift strategy, and resonating with previous single unit recording findings in monkey prefrontal cortex. Our results highlight both similarities and differences between exploration behavior as it emerges in meta-learning RNNs, and computational mechanisms identified in cognitive and systems neuroscience work." @default.
- W4367172263 created "2023-04-28" @default.
- W4367172263 creator A5032845560 @default.
- W4367172263 creator A5051054354 @default.
- W4367172263 creator A5091662459 @default.
- W4367172263 date "2023-04-27" @default.
- W4367172263 modified "2023-10-16" @default.
- W4367172263 title "Human-level reinforcement learning performance of recurrent neural networks is linked to hyperperseveration, not directed exploration" @default.
- W4367172263 cites W1760547568 @default.
- W4367172263 cites W1860349330 @default.
- W4367172263 cites W1980324747 @default.
- W4367172263 cites W1981273464 @default.
- W4367172263 cites W1993411524 @default.
- W4367172263 cites W2000702984 @default.
- W4367172263 cites W2001643038 @default.
- W4367172263 cites W2001948633 @default.
- W4367172263 cites W2017539895 @default.
- W4367172263 cites W2019310225 @default.
- W4367172263 cites W2021520260 @default.
- W4367172263 cites W2027838193 @default.
- W4367172263 cites W2036783885 @default.
- W4367172263 cites W2044924321 @default.
- W4367172263 cites W2047125104 @default.
- W4367172263 cites W2054698589 @default.
- W4367172263 cites W2058279623 @default.
- W4367172263 cites W2064675550 @default.
- W4367172263 cites W2077611535 @default.
- W4367172263 cites W2081616632 @default.
- W4367172263 cites W2083058253 @default.
- W4367172263 cites W2084489534 @default.
- W4367172263 cites W2105934661 @default.
- W4367172263 cites W2107886772 @default.
- W4367172263 cites W2110123411 @default.
- W4367172263 cites W2111534506 @default.
- W4367172263 cites W2124136621 @default.
- W4367172263 cites W2145339207 @default.
- W4367172263 cites W2147971214 @default.
- W4367172263 cites W2148534890 @default.
- W4367172263 cites W2153554101 @default.
- W4367172263 cites W2168405694 @default.
- W4367172263 cites W2170638884 @default.
- W4367172263 cites W2194321275 @default.
- W4367172263 cites W2203714058 @default.
- W4367172263 cites W2214962513 @default.
- W4367172263 cites W2276520418 @default.
- W4367172263 cites W2323363676 @default.
- W4367172263 cites W2376518725 @default.
- W4367172263 cites W2536957931 @default.
- W4367172263 cites W2547503455 @default.
- W4367172263 cites W2558230882 @default.
- W4367172263 cites W2566424235 @default.
- W4367172263 cites W2766447205 @default.
- W4367172263 cites W2778853303 @default.
- W4367172263 cites W2790414966 @default.
- W4367172263 cites W2883084823 @default.
- W4367172263 cites W2900107753 @default.
- W4367172263 cites W2902847783 @default.
- W4367172263 cites W2902907165 @default.
- W4367172263 cites W2919115771 @default.
- W4367172263 cites W2922516504 @default.
- W4367172263 cites W2938321354 @default.
- W4367172263 cites W2944997976 @default.
- W4367172263 cites W2949369413 @default.
- W4367172263 cites W2949436289 @default.
- W4367172263 cites W2949941589 @default.
- W4367172263 cites W2981588505 @default.
- W4367172263 cites W2983405298 @default.
- W4367172263 cites W3000294554 @default.
- W4367172263 cites W3006117593 @default.
- W4367172263 cites W3016391357 @default.
- W4367172263 cites W3021727781 @default.
- W4367172263 cites W3030493409 @default.
- W4367172263 cites W3034751418 @default.
- W4367172263 cites W3041566370 @default.
- W4367172263 cites W3042790281 @default.
- W4367172263 cites W3080525189 @default.
- W4367172263 cites W3095203791 @default.
- W4367172263 cites W3096511324 @default.
- W4367172263 cites W3119012401 @default.
- W4367172263 cites W3129154534 @default.
- W4367172263 cites W3138307461 @default.
- W4367172263 cites W3153977456 @default.
- W4367172263 cites W3174205044 @default.
- W4367172263 cites W3204059462 @default.
- W4367172263 cites W4205787442 @default.
- W4367172263 cites W4220747123 @default.
- W4367172263 cites W4220823496 @default.
- W4367172263 cites W4226243011 @default.
- W4367172263 cites W4247460979 @default.
- W4367172263 cites W4254816979 @default.
- W4367172263 cites W4283271670 @default.
- W4367172263 cites W4283760982 @default.
- W4367172263 cites W4294629941 @default.
- W4367172263 cites W4297256711 @default.
- W4367172263 cites W4297730339 @default.
- W4367172263 cites W4309267673 @default.
- W4367172263 cites W4311663566 @default.
- W4367172263 doi "https://doi.org/10.1101/2023.04.27.538570" @default.
- W4367172263 hasPublicationYear "2023" @default.
- W4367172263 type Work @default.