Neural Circuits for Fast Poisson Compressed Sensing in the Olfactory Bulb

Area of Science:

• Neuroscience
• Computational Biology
• Sensory Systems

Background:

• Mammalian olfactory systems process complex odor information from noisy inputs within a single sniff.
• Existing compressed sensing models for olfaction lack the anatomical and physiological specificity of the olfactory bulb and fail to meet timescale constraints.

Purpose of the Study:

• To propose a novel rate-based Poisson compressed sensing circuit model for the olfactory bulb.
• To investigate if this model can explain fast and accurate odor decoding within the timescale of a sniff.
• To explore the model's capacity for uncertainty estimation and its relationship to neural coding geometry.

Main Methods:

• Developed a rate-based Poisson compressed sensing circuit model incorporating olfactory bulb neuron classes, connectivity, and physiology.
• Simulated the model with circuit sizes comparable to the human olfactory bulb.
• Analyzed the model's performance in odor detection, concentration estimation, and Bayesian posterior sampling.

Main Results:

• The proposed model accurately detects tens of odors within the 100-millisecond timescale of a single sniff.
• The model successfully performs Bayesian posterior sampling for uncertainty estimation.
• Fast inference is achieved when the neural code geometry aligns with receptor properties, resulting in a distributed, non-axis-aligned code.

Conclusions:

• Normative modeling of the olfactory bulb circuit can explain fast and accurate odor perception.
• The model provides a framework for mapping olfactory bulb function to its specific neural architecture.
• Results suggest that the geometry of the neural code is critical for efficient olfactory processing and uncertainty estimation.