Tuning positive feedback for signal detection in noisy dynamic environments

Area of Science:

• Computational neuroscience
• Decision-making algorithms
• Biological systems modeling

Background:

• Decision-making relies on learning from past actions.
• Biological systems like neuronal assemblies and insect societies integrate positive feedback and memory decay.
• These systems process signals amidst noise, crucial for survival and function.

Purpose of the Study:

• To investigate how biological systems handle dynamic signal detection in noisy environments.
• To analyze the role of positive feedback and memory decay in processing weak signals.
• To evaluate the efficiency of a simple positive feedback algorithm in tracking dynamic inputs.

Main Methods:

• Modeling a dynamic two-armed bandit problem.
• Analyzing the interplay between positive feedback strength and decay rate.
• Applying signal detection theory and Fisher efficiency metrics.

Main Results:

• A single tracking variable, with tuned feedback and decay, can detect dynamic inputs amid significant noise.
• The proposed positive feedback algorithm demonstrates Fisher efficiency.
• The algorithm's tracking time is dependent on signal magnitude (|h|) and noise magnitude (σ).

Conclusions:

• Optimized positive feedback and forgetting mechanisms are key for robust signal detection in biological systems.
• Simple algorithms, when appropriately tuned, can achieve high efficiency in complex dynamic environments.
• This framework provides insights into efficient information processing in both biological and artificial systems.