From Tutorial on Neural Systems Modeling, Chapter 7, pg. 238: We will refer to the upper motoneurons that elicit "spin" and "freeze" responses as the spin upper-motoneuron (SUMO) and the freeze upper-motoneuron (FUMO), respectively. We will assume that both SUMO and FUMO receive auditory input and respond to tone. We will further assume that the behavioral response of the rabbit will be determined by whichever upper-motoneuron schema has the larger response to the tone. For example, if the response of SUMO is greater than that of FUMO, then SUMO will win the competition for control of the behavioral response, and the rabbit will spin the wheel rather than freeze. However, since there is an element of randomness in behavior, we will assume that the rabbit will sometimes innovate (i.e., explore). This means that the rabbit will sometimes produce the behavior represented by the upper-motoneuron schema that loses the competition for behavioral control. Finally, we assume that the rabbit can learn to change its behavioral response to the tone by adjusting the sizes of the responses of SUMO and FUMO to the tone. We assume that this learning is driven by reinforcement signals, and that the overall rabbit model expresses reinforcement learning by acquiring a simu­lated version of the avoidance response. The ability of the model to simulate this form of reinforcement learning depends entirely on its structure. The model will incorporate a HEAR schema that represents the auditory system. We will design the model so that the weights of the excitatory connec­tions from the HEAR schema to either upper-motoneuron schema are modifi­able. On each trial, the weight that is modified will be the one associated with the connection from HEAR to whichever upper motoneuron represented the behavioral response produced by the simulated rabbit. For example, if the rab­bit spins the wheel, then the weight of the connection to SUMO from HEAR will be modified. The amount of this modification will be proportional to the difference between a reinforcement signal and the current weight value. Thus the weights from HEAR, and so the responses of SUMO and FUMO to the tone, will track the reinforcement associated with a spin or a freeze response, respectively.

NanoBio Node, University of Illinois Champaign-Urbana

Researchers should cite this work as follows:

• Tutorial on Neural Systems Modeling, Copyright 2010 Sinauer Associates Inc. Author: Thomas J. Anastasio

• AbderRahman N Sobh (2014), "[Illinois]: Avoidance Learn Simulation," https://nanohub.org/resources/avoidlearn. (DOI: 10.4231/D3HD7NS55).

  BibTex | EndNote

1. Illinois
2. MCB 493
3. nano/bio
4. NanoBio Node at Illinois
5. Neural Systems Modeling