Three-Factor Learning in Spiking Neural Networks: An Overview of Methods and Trends from a Machine Learning Perspective

TL;DR

This paper surveys three-factor learning in Spiking Neural Networks from a machine learning perspective, detailing how a modulatory signal g_t augments local STDP to enable reward-guided and context-aware plasticity. It synthesizes theoretical foundations, including Δw_t = H(t_{pre}, t_{post}, g_t) and its intrinsic/extrinsic modulation, with algorithmic developments such as eprop, MDGL, and ETLP, as well as datasets, domains, and hardware platforms. Key contributions include a structured overview of learning rules, neuromodulatory mechanisms, scalability challenges, and encoding strategies, along with a critical view of limitations and gaps. The work highlights cross-disciplinary opportunities to validate models with real-world data and neuromorphic hardware, aiming to accelerate energy-efficient, adaptive AI systems for robotics and cognitive computing.

Abstract

Three-factor learning rules in Spiking Neural Networks (SNNs) have emerged as a crucial extension to traditional Hebbian learning and Spike-Timing-Dependent Plasticity (STDP), incorporating neuromodulatory signals to improve adaptation and learning efficiency. These mechanisms enhance biological plausibility and facilitate improved credit assignment in artificial neural systems. This paper takes a view on this topic from a machine learning perspective, providing an overview of recent advances in three-factor learning, discusses theoretical foundations, algorithmic implementations, and their relevance to reinforcement learning and neuromorphic computing. In addition, we explore interdisciplinary approaches, scalability challenges, and potential applications in robotics, cognitive modeling, and AI systems. Finally, we highlight key research gaps and propose future directions for bridging the gap between neuroscience and artificial intelligence.

Figures (8)

• Figure 1: Spiking Neural Network domain and its key components. In our work, we focus explicitly on the highlighted subdomain - machine learning.
• Figure 2: General overview of three-factor learning principle: synaptic weights are modified based on local activity and influence of the third factor.
• Figure 3: Possible influences of third factor on Spike-Timing Dependent Plasticity (STDP) learning rule. The A plot shows baseline STDP, where synaptic weight change ($\Delta w$) depends on the relative timing ($\Delta t$) of pre- and post-synaptic spikes. The B plot illustrates reversed STDP, where the LTD and LTP polarities are flipped. The C plot demonstrates STDP shape modulation, where neuromodulatory factors influence the temporal profile of plasticity, modifying the learning window width.
• Figure 4: Different sources of the top-level third factor: the signal can be emitted intrinsically between neurons in the same neuronal circuit or extrinsically, when the signal arrives from outside of the circuit. Neurons that are coinciding either spatially, functionally or morphologically are considered to be in the same circuit Marder2002.
• Figure 5: Datasets used in research papers investigating three-factor learning in SNNs.