Abrupt, Asynchronous Changes in Action Representations by Anterior Cingulate Cortex Neurons during Trial and Error Learning

A single fast Hebbian-like process enabling one-shot class addition in deep neural networks without backbone modification

One-shot learning, the ability to learn a new concept from a single instance, is a distinctive brain function that has garnered substantial interest in machine learning. While modeling physiological mechanisms poses challenges, advancements in artificial neural networks have led to performances in specific tasks that rival human capabilities. Proposing one-shot learning methods with these advancements, especially those involving simple mechanisms, not only enhance technological development but also contribute to neuroscience by proposing functionally valid hypotheses. Among the simplest methods for one-shot class addition with deep learning image classifiers is "weight imprinting," which uses neural activity from a new class image data as the corresponding new synaptic weights. Despite its simplicity, its relevance to neuroscience is ambiguous, and it often interferes with original image classification, which is a significant drawback in practical applications. This study introduces a novel interpretation where a part of the weight imprinting process aligns with the Hebbian rule. We show that a single Hebbian-like process enables pre-trained deep learning image classifiers to perform one-shot class addition without any modification to the original classifier's backbone. Using non-parametric normalization to mimic brain's fast Hebbian plasticity significantly reduces the interference observed in previous methods. Our method is one of the simplest and most practical for one-shot class addition tasks, and its reliance on a single fast Hebbian-like process contributes valuable insights to neuroscience hypotheses.

From ensembles to meta-ensembles: Specific reward encoding by correlated network activity

Neuronal ensembles are local, sparsely distributed populations of neurons that are reliably re-activated by a specific stimulus, context or task. Such discrete cell populations can be defined either functionally, by electrophysiological recordings or in vivo calcium imaging, or anatomically, using the expression of markers such as the immediate early gene cFos. A typical example of tasks that involve the formation of neuronal ensembles is reward learning, such as the cue-reward pairing during operant conditioning. These ensembles are re-activated during cue-presentation and increasing evidence suggests that this re-activation is the neurophysiological basis for the execution of reward-seeking behavior. Whilst the pursuit of rewards is a common daily activity, it is also related to the consumption of drugs, such as alcohol, and may result in problematic behaviors including addiction. Recent research has identified neuronal ensembles in several reward-related brain regions that control distinct aspects of a conditioned response, e.g., contextual information about the availability of a specific reward or the actions needed to retrieve this reward under the given circumstances. Here, we review studies using the activity marker cFos to identify and characterize neuronal ensembles related to alcohol and non-drug rewards with a special emphasis on the discrimination between different rewards by meta-ensembles, i.e., by dynamic co-activation of multiple ensembles across different brain areas.

Event-based control of autonomic and emotional states by the anterior cingulate cortex

Despite being an intensive area of research, the function of the anterior cingulate cortex (ACC) remains somewhat of a mystery. Human imaging studies implicate the ACC in various cognitive functions, yet surgical ACC lesions used to treat emotional disorders have minimal lasting effects on cognition. An alternative view is that ACC regulates autonomic states, consistent with its interconnectivity with autonomic control regions and that stimulation evokes changes in autonomic/emotional states. At the cellular level, ACC neurons are highly multi-modal and promiscuous, and can represent a staggering array of task events. These neurons nevertheless combine to produce highly event-specific ensemble patterns that likely alter activity in downstream regions controlling emotional and autonomic tone. Since neuromodulators regulate the strength of the ensemble activity patterns, they would regulate the impact these patterns have on downstream targets. Through these mechanisms, the ACC may determine how strongly to react to the very events its ensembles represent. Pathologies arise when specific event-related representations gain excessive control over autonomic/emotional states.