Issues in data expansion in understanding criticality in biological systems

Area of Science:

• Neuroscience
• Statistical Physics
• Complex Systems

Background:

• Second-order phase transitions, or critical points, exhibit scale-invariance and universality, independent of microscopic details.
• Biological systems displaying similar emergent properties despite diverse microscopic components are hypothesized to operate near criticality.
• Advancements in neuroscience enable the investigation of criticality in neural networks, but dataset limitations pose challenges.

Purpose of the Study:

• To develop an analytical method for expanding finite-sized neural network datasets to the large N limit.
• To assess the impact of different dataset expansion strategies on criticality analysis.
• To evaluate the reliability of current methods for inferring criticality in biological systems.

Main Methods:

• Development of an analytical method for dataset expansion to the large N limit.
• Application of the method to finite-sized neural network datasets.
• Comparative analysis of criticality measures under different expansion protocols.

Main Results:

• The study demonstrates that distinct dataset expansion techniques, while preserving mean and variance, produce divergent conclusions regarding criticality.
• This variability casts doubt on the robustness of established procedures for deducing criticality from finite datasets.
• The findings highlight the sensitivity of criticality assessments to the chosen data expansion methodology.

Conclusions:

• Established methods for inferring criticality in biological systems by expanding finite datasets may be unreliable.
• The choice of dataset expansion method significantly influences the assessment of criticality.
• Further research is needed to develop more robust and reliable methods for studying criticality in neural networks and other biological systems.