Relationship between US response and associated compartment readout.

Results are shown for a representative example optimized network as in Fig 3. Each point represents one compartment. The horizontal axis denotes the corresponding entry of Wreadout and the vertical axis denotes the difference between that compartment’s dopamine neuron’s response to a positive and negative valence US. Consistent with learning involving depression of KC-to-MBON synapses, compartments whose output neuron activity biases the readout toward negative valence (avoidance) have dopamine neurons that preferentially respond to positive valence US. (EPS)