Mixed signal learning by spike correlation propagation in feedback inhibitory circuits

Area of Science:

• Computational Neuroscience
• Neural Networks
• Synaptic Plasticity

Background:

• The brain effectively processes mixed signals despite noise.
• Spike-timing-dependent plasticity (STDP) is a key synaptic learning mechanism.
• Lateral circuits and feedback connections in the brain generate secondary spike correlations, but their role in STDP learning remains unclear.

Purpose of the Study:

• To investigate how propagated correlations in lateral circuits influence STDP learning.
• To determine if these correlations benefit learning from uncertain stimuli.
• To understand the learning algorithms implemented by such circuits.

Main Methods:

• Modeling feedforward networks with lateral inhibitory circuits.
• Analyzing the influence of propagated correlation on STDP.
• Deriving analytical conditions for signal detection via STDP.
• Applying dynamical systems and information theory.

Main Results:

• Analytical conditions for STDP-based minor signal detection were derived.
• Different correlation timescales are beneficial depending on noise type; non-precise correlation aids learning with cross-talk noise.
• Excitatory and inhibitory STDP at lateral connections optimize circuits for signal detection.
• The model performs blind source separation, akin to Bayesian independent component analysis.

Conclusions:

• Propagated correlations in lateral circuits play a crucial role in STDP learning.
• The brain's feedback circuits can adapt for optimal signal detection and noise handling.
• This work provides fundamental insights into STDP learning within complex neural architectures.