This thesis makes two major contributions: it introduces a novel method for analysis of artificial neural networks and provides new models of the nematode Caenorhabditis elegans nervous system. The analysis method extracts neural network motifs, or subnetworks of recurring neuronal function, from optimized neural networks. The method first creates models for each neuron relating network stimulus to neuronal response, then clusters the model parameters, and finally combines the neurons into multi-neuron motifs based on their cluster category. To infer biological function, this analysis method was applied to neural networks optimized to reproduce C. elegans behavior, which converged upon a small number of motifs. This allowed both a quantitative exploration of network function as well as discovery of larger motifs.

Neural network models of C. elegans anatomical connectivity were optimized to reproduce two C. elegans behaviors: chemotaxis (orientation towards a maximum chemical attractant concentration) and thermotaxis (orientation towards a set temperature). Three chemotaxis motifs were identified. Experimental evidence suggests that chemotaxis is driven by a differentiator motif with two important features. The first feature was a fast, excitatory pathway in parallel with one or more slow, inhibitory pathways. The second feature was inhibitory feedback on all self-connections and recurrent loops, which regulates neuronal response. Six thermotaxis motifs were identified. Every motif consisted of two circuits, each a previously discovered chemotaxis motif with most having a dedicated sensory neuron. One circuit was thermophilic (heat-seeking) and the other was cryophilic (cold-seeking). Experimental evidence suggests that the cryophilic circuit is a differentiator motif and the thermophilic circuit functions by klinokinesis.