腦神經系統如何建立身體動作相關的內在模型

An algorithm of how nervous system establishes the internal model of body movement

薛又齊(Sik, Pan Shee)

指導教授 ： 羅中泉 ； 王道維

國立清華大學/理學院/物理學系/碩士(2020年)

摘要

任何身體動作都伴隨着感知回饋。有一種假設表示神經系統建構了一個內在模型，可以預測動作帶來的感知回饋，並且將此預測與對外界的實際觀測比較，來達到控管動作的性能的效果。但神經系統是如何學習並建構如此的內在模型，至今還是一個未知的謎題。爲了解開這個問題，我們提出了一個基於果蠅的中央復合體 (Central Complex, CX) 中存在的橢圓體腦區 (Ellipsoid Body, EB) 和前腦橋 (Protocerebral Bridge, PB)，是對外界視覺刺激反應，並且可以形成及維持空間記憶的一個腦區迴路。我們的模型包含了兩個主要環形迴路：一個對視覺刺激反應，反映出外界的世界，另一個接受動作感知回饋，達到預測的效果，如此形成一個內在模型。兩個環形迴路連接至一個稱爲差錯偵測單位 (mismatch detector) 的小型迴路，是可以在偵測到輸入差異存在時輸出訊號的迴路。然後假設了一個學習公式，透過這個公式表現內在模型如何自動透過視覺經驗建構起來，以及動作感知的對應關系如何透過此過程變化。我們提出的模型提供神經系統一個可靠的學習機制來解釋內在模型的功用。

關鍵字

內在模型 ； 果蠅 ； 橢圓體 ； 中央複合體 ； 差錯偵測 ； 預測編碼 ； 計算神經模型

並列摘要

Any body movement is associated with sensory feedback. It is hypothesized that the nervous system establishes an internal model that can predict the feedback and compare the prediction with the observation in order to monitor the performance of the movement. However, it is not clear how a nervous system learns to establish such an internal model. To address this issue, we propose a neural network model based on the Ellipsoid Body (EB) and Protocerebral Bridge (PB) system found in Drosophila Melanogaster. EB and PB are neuropils in the Central Complex (CX), which is known to respond to visual stimuli and maintain an activity bump that represents the orientation of the stimulus, or the relative head direction. Our network model consists of two ring circuits: one responds to the visual input and represents the external world; the other receives motor feedback (corollary discharge) and forms an internal model. The two rings are connected with the mismatch detectors which signal mismatch between the two representations. We include a learning mechanism and show how an internal model can be automatically established through the visual experience and how a new internal model can gradually form when the "motor-sensory" mapping is changed. Our neural circuit model provides a plausible neural mechanism for the internal model.

並列關鍵字

Internal model ； Drosophila ； Ellipsoid Body ； Central Complex ； Mismatch detection ； Predictive coding ； Computational neuroscience model

參考文獻

1. Land MF. Do we have an internal model of the outside world?. Philos Trans R Soc Lond B Biol Sci. (2014).

2. Seelig, J. D. & Jayaraman, V. Neural dynamics for landmark orientation and angular path integration. Nature 521, 186–191 (2015).

3. Green, J. et al. A neural circuit architecture for angular integration in Drosophila. Nature 546, 101–106 (2017).

4. Turner-Evans, D. et al. Angular velocity integration in a fly heading circuit. Elife 6, 1–39 (2017).

5. Taube, J. S., Muller, R. U. & Ranck, J. B. Head-direction cells recorded from the postsubiculum in freely moving rats. I. Description and quantitative analysis. J. Neurosci. 10, 420–435 (1990).