Neuromodulatory systems such as serotonin and dopamine seem to play an important role in our ability to effortlessly adapt to and learn in unpredictable, ever-changing environments. These systems can dynamically change how neurons learn and respond to stimuli based on context. Our brain has a diverse set of neuromodulatory systems, operating at multiple time scales and interacting in complex ways.
Can neuromodulatory principles enable better and more flexible learning in deep neural networks?
In this Brains@Bay meetup, our speakers reviewed what is known about neuromodulatory mechanisms and explored how they may provide useful insights into creating more efficient and truly intelligent machines.
➤ Srikanth Ramaswamy (Newcastle University)
➤ Jie Mei (The Brain and Mind Institute)
➤ Thomas Miconi (ML Collective)
The talks were followed by a discussion panel and Q&A.
Meetup link: https://www.meetup.com/Brains-Bay/events/282182527/
Brains@Bay Meetups focus on how neuroscience can inspire us to create improved artificial intelligence and machine learning algorithms. Find more details here.
Links mentioned during the Meetup:
- For a very comprehensive review of neuromodulators and their roles in health and disease, see https://www.frontiersin.org/articles/10.3389/fncir.2017.00108/full
- Neuromodulators generate multiple context-relevant behaviors in a recurrent neural network by shifting activity hypertubes: https://www.biorxiv.org/content/10.1101/2021.05.31.446462v2.abstract
Srikanth Ramaswamy, A primer on neuromodulatory systems;
Jie Mei, Implementing multi-scale neuromodulation in artificial networks
Abstract: Neuromodulators are signalling chemicals in the brain, which control the emergence of adaptive learning and behaviour. Neuromodulators including dopamine, acetylcholine, serotonin and noradrenaline operate on a spectrum of spatio-temporal scales in tandem and opposition to reconfigure functions of biological neural networks and to regulate global cognition and state transition. Although neuromodulators are important in shaping cognition, their phenomenology is yet to be fully realized in deep neural networks (DNNs). In this talk, we will first give an overview of the biological organizing principles of neuromodulators in adaptive cognition and highlight the competition and cooperation across neuromodulators. We will then discuss ongoing research on bio-inspired mechanisms of neuromodulatory function in DNNs, and propose a computational framework to incorporate their diverse functional settings and inspire new architectures of “neuromodulation-aware” DNNs.
Thomas Miconi, How to evolve your own lab rat: Learning to acquire novel cognitive tasks with evolution, plasticity and meta-meta-learning
Abstract: A hallmark of intelligence is the ability to learn new flexible, cognitive behaviors – that is, behaviors that require discovering, storing and exploiting novel information for each new instance of the task. In meta-learning, agents are trained with external algorithms to learn one specific cognitive task. However, animals are able to pick up such cognitive tasks automatically, as a result of their evolved neural architecture and synaptic plasticity mechanisms, including neuromodulation. Here we evolve neural networks, endowed with plastic connections and reward-based neuromodulation, over a sizable set of simple meta-learning tasks based on a framework from computational neuroscience. The resulting evolved networks can automatically acquire a novel simple cognitive task, never seen during evolution, through the spontaneous operation of their evolved neural organization and plasticity system. We suggest that attending to the multiplicity of loops involved in natural learning may provide useful insight into the emergence of intelligent behavior.
- Thomas Miconi. Learning to acquire novel cognitive tasks with evolution, plasticity and meta-meta-learning, arXiv, 2021