The goal of spiking network modeling for our Animat project is trying to find the mapping between network activity and motor control for the robot, and the mapping between sensory input to the robot and the feedback stimulation patterns. Those two mappings connect neurons to the robot to form a closed-loop control system. We focus on the mappings that enable the system to show “learning” behaviors. The definition of “learning” here is the change of goal-directed behavior to show the improvement of performance.
Obviously, the advantages of modeling are that the internal properties of the network can be monitored with arbitrary spatial and temporal resolutions, and the simulation can be setup easily and proceed fast. We use modeling to make predictions (especially on mappings) for our experimental designs when working with the real biological network.
Method
A standard leaky-integrate-and-fire neuron model is implemented. The neuronal network consists of 1000 LIF neurons with 10000-50000 synapses that include long-range connections, with 30% inhibitory neurons and 30% self-firing neurons. Frequency-dependent dynamic synapses were introduced for modeling synaptic depression.
Movies
1) new_connection.avi (103 Mb):The colored circles indicate the locations of 1000 LIF neurons. The color codes represent the membrane potential which corresponding values are shown in color bar. This movie is replayed 10X slower.
2) new_connection_spk.avi (50 Mb):The same simulation as previous movie, but is replayed 50X slower. The blue circles indicate the locations of 1000 LIF neurons. And the processes of firing neurons are shown in blue lines.
References
- Chao, Z. C., Bakkum, D. J., Wagenaar, D. A. and Potter, S. M. (2005) Effects of Random External Background Stimulation on Network Synaptic Stability After Tetanization: A Modeling Study . Neuroinformatics 3(3): 263-280. Download
- Chao, Z. C., Bakkum, D. J., & Potter, S. M. (2007). Region-specific network plasticity in simulated and living cortical networks: Comparison of the center of activity trajectory (CAT) with other statistics. Journal of Neural Engineering, 4, 294-308. Download