CHANI: Correlation-based Hawkes Aggregation of Neurons with bio-Inspiration

2025-12-31 · Source: JMLR · Field: Technology & Digital — Artificial Intelligence & Machine Learning · Depth: Expert, quick

Summary

The CHANI (Correlation-based Hawkes Aggregation of Neurons with bio-Inspiration) spiking neural network, detailed in a 2026 paper by Jaffard, Vaiter, and Reynaud-Bouret, demonstrates a biologically inspired approach to classification learning using only local transformations. This network models neuron activity with Hawkes processes and updates synaptic weights via an expert aggregation algorithm, ensuring a local and simple learning rule. The authors mathematically proved that CHANI can learn both on average and asymptotically. Furthermore, the network automatically forms neuronal assemblies, allowing it to encode multiple classes where intermediate layer neurons can be activated by more than one class. Numerical simulations on synthetic datasets support these theoretical findings, offering a contrast to typical empirical validations of bio-inspired networks.

Key takeaway

For AI scientists exploring biologically inspired neural networks, CHANI offers a mathematically validated framework for local learning and neuronal assembly formation. You should consider its Hawkes process modeling and expert aggregation algorithm as a robust theoretical foundation for developing SNNs capable of complex classification tasks, moving beyond purely empirical validation.

Key insights

CHANI is a bio-inspired spiking neural network mathematically proven to learn classification tasks via local rules.

Principles

• Local learning rules can enable complex concept representation.
• Hawkes processes effectively model neuron activity.
• Expert aggregation updates synaptic weights simply.

Method

CHANI models neuron activity with Hawkes processes. Synaptic weights are updated using an expert aggregation algorithm, providing a local learning rule that facilitates classification and neuronal assembly formation.

In practice

• Implement Hawkes processes for SNN neuron modeling.
• Utilize expert aggregation for local weight updates.
• Explore SNNs for multi-class encoding.

Topics

• CHANI
• Spiking Neural Networks
• Hawkes Processes
• Expert Aggregation Algorithm
• Neuronal Assemblies

Code references

• SophieJaffard/CHANI
• JmlrOrg/v27

Best for: AI Scientist, Research Scientist

Related on AIssential

• Explainable AI based ensemble model for the identification of Schizophrenia prodromal phase — An XAI-integrated ensemble ML model accurately predicts Schizophrenia prodromal symptoms for early clinical intervention.
• Harnessing cortical geometry, wiring, and function as inductive biases for recurrent neural networks — Cortical geometry, wiring, and function serve as powerful inductive biases for more effective recurrent neural networks.
• Pangu-ACE: Adaptive Cascaded Experts for Educational Response Generation on EduBench — Adaptive cascaded expert systems can improve quality and validity in educational response generation by selectively routing tasks.
• Structure as Computation: Developmental Generation of Minimal Neural Circuits — Biological developmental rules sculpt neural circuits that are highly amenable to rapid, efficient learning.
• HawkesNest: A Multi-Axis Synthetic Benchmark for Spatiotemporal Pattern Complexity — HawkesNest provides a multi-axis synthetic benchmark for diagnostic stress testing of spatiotemporal point process models.
• Amortized Inference of Neuron Parameters on Analog Neuromorphic Hardware — Amortized SBI effectively parameterizes analog neuron circuits on neuromorphic hardware, with summary networks outperforming handcrafted statistics.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by JMLR.