Inhibitory STDP generates inverse models through detailed balance

摘要: In the song bird ('backward') mappings from sensory representations to motor areas recently were proposed that would 'postdict' activations during singing. Such a sensor-motor mapping represents an inverse model of motor-sensor-loop passing through world and thereby can explain impressive imitation capabilities birds [1]. The neurobiological mechanisms might generate, fine tune continuously adapt such models, however, are not known. Here we show spike timing dependent plasticity (STDP) inhibitory synapses is sufficient for self-organisation in simple closed loop motor-sensor-motor system [2]. Similar case forward where predictable, self-generated inputs become suppressed [3], mechanism generates sparse activities by cancelling predictable fluctuations neurons' excitabilities. Our results be learned with elementary biologically plausible learning rule thus could underly learning. our presentation will discuss also potential relevance this operation state recurrent networks as e.g. cortex [4].