選擇性抑制型神經元對於吸引子決策網路的影響

在危急的情況下，能夠準確且快速地做出判斷是對於生物體的生存而言至關重要的能力。過去的許多研究利用Random dots motion task 來研究準確度與速度之間的取捨(trade-off)。其中，吸引子網路能夠成功的解釋實驗數據，且其所預測的準確性與反應時間之平均值與實驗數據吻合。然而，新的實驗數據與吸引子模型的其中一項假設不吻合。此實驗數據顯示決策模型中的抑制型神經應與刺激型神經擁有同樣程度的選擇性，然而吸引子模型用的卻是沒有選擇性的抑制型神經。因此，在此研究中我們修正了此一假設，並且比較修正前後的模型之間的計算能力的差別。我們用了一個簡化版的模型來研究這兩個模型，並且發現只有有選擇性的模型才能產生一個隨時變的位能井。時變性讓選擇性模型得以在初期時小心衡量資訊，並在後期時快速做出決定。因此，選擇性模型得以做出比無選擇性模型更快且更準確的決定。

關鍵字

選擇性；吸引子網路；決策；位能井

並列摘要

The ability to decide swiftly and accurately in an urgent scenario is crucial for an organism’s survival. Past studies have used the random dots motion task to quantify the speed-accuracy trade-off that plagues most decision making tasks, and the attractor model is a model that has successfully captured the average accuracy and reaction time as seen in the experimental data. However, new experimental findings call into question the validity of the attractor model’s network configuration. In particular, it is shown that the inhibitory neurons in the posterior parietal cortex of mice are as selective to decision making results as the excitatory neurons, which is a direct contradiction to the attractor model’s assumption of unselective inhibition. In this study, we investigate a revised model of the classic attractor model, and analyze in detail the differences in computational ability that it brings. We proposed a reduced model for both the unselective and selective models, and showed that the selective model alone produces a time-varying energy landscape. This time dependence of the energy landscape allows the selective model to integrate information carefully in initial stages, then quickly converge to an attractor once the choice is clear. This results in the selective model having a more efficient speed-accuracy trade-off that is unreachable by unselective models.