Learning and generalization of within-category representations in a rule-based category structure

Stable, flexible, common, and distinct behaviors support rule-based and information-integration category learning

The ability to organize variable sensory signals into discrete categories is a fundamental process in human cognition thought to underlie many real-world learning problems. Decades of research suggests that two learning systems may support category learning and that categories with different distributional structures (rule-based, information-integration) optimally rely on different learning systems. However, it remains unclear how the same individual learns these different categories and whether the behaviors that support learning success are common or distinct across different categories. In two experiments, we investigate learning and develop a taxonomy of learning behaviors to investigate which behaviors are stable or flexible as the same individual learns rule-based and information-integration categories and which behaviors are common or distinct to learning success for these different types of categories. We found that some learning behaviors are stable in an individual across category learning tasks (learning success, strategy consistency), while others are flexibly task-modulated (learning speed, strategy, stability). Further, success in rule-based and information-integration category learning was supported by both common (faster learning speeds, higher working memory ability) and distinct factors (learning strategies, strategy consistency). Overall, these results demonstrate that even with highly similar categories and identical training tasks, individuals dynamically adjust some behaviors to fit the task and success in learning different kinds of categories is supported by both common and distinct factors. These results illustrate a need for theoretical perspectives of category learning to include nuances of behavior at the level of an individual learner.

Category learning in a recurrent neural network with reinforcement learning

It is known that humans and animals can learn and utilize category information quickly and efficiently to adapt to changing environments, and several brain areas are involved in learning and encoding category information. However, it is unclear that how the brain system learns and forms categorical representations from the view of neural circuits. In order to investigate this issue from the network level, we combine a recurrent neural network with reinforcement learning to construct a deep reinforcement learning model to demonstrate how the category is learned and represented in the network. The model consists of a policy network and a value network. The policy network is responsible for updating the policy to choose actions, while the value network is responsible for evaluating the action to predict rewards. The agent learns dynamically through the information interaction between the policy network and the value network. This model was trained to learn six stimulus-stimulus associative chains in a sequential paired-association task that was learned by the monkey. The simulated results demonstrated that our model was able to learn the stimulus-stimulus associative chains, and successfully reproduced the similar behavior of the monkey performing the same task. Two types of neurons were found in this model: one type primarily encoded identity information about individual stimuli; the other type mainly encoded category information of associated stimuli in one chain. The two types of activity-patterns were also observed in the primate prefrontal cortex after the monkey learned the same task. Furthermore, the ability of these two types of neurons to encode stimulus or category information was enhanced during this model was learning the task. Our results suggest that the neurons in the recurrent neural network have the ability to form categorical representations through deep reinforcement learning during learning stimulus-stimulus associations. It might provide a new approach for understanding neuronal mechanisms underlying how the prefrontal cortex learns and encodes category information.

The impact of training methodology and representation on rule-based categorization: An fMRI study

Hélie, Shamloo, & Ell (2017) showed that regular classification learning instructions (A/B) promote between-category knowledge in rule-based categorization whereas conceptual learning instructions (YES/NO) promote learning within-category knowledge with the same categories. Here we explore how these tasks affect brain activity using fMRI. Participants learned two sets of two categories. Computational models were fit to the behavioral data to determine the type of knowledge learned by each participant. fMRI contrasts were computed to compare BOLD signal between the tasks and between the types of knowledge. The results show that participants in the YES/NO task had more activity in the pre-supplementary motor area, prefrontal cortex, and the angular/supramarginal gyrus. These brain areas are related to working memory and part of the dorsal attention network, which showed increased task-based functional connectivity with the medial temporal lobes. In contrast, participants in the A/B task had more activity in the thalamus and caudate. These results suggest that participants in the YES/NO task used bivalent rules and may have treated each contextual question as a separate task, switching task each time the question changed. Activity in the A/B condition was more consistent with participants applying direct Stimulus → Response rules. With regards to knowledge representation, there was a large shared network of brain areas, but participants learning between-category information showed additional posterior parietal activity, which may be related to the inhibition of incorrect motor programs.

A Study of Individual Differences in Categorization with Redundancy

Humans and other animals are constantly learning new categories and making categorization decisions in their everyday life. However, different individuals may focus on different information when learning categories, which can impact the category representation and the information that is used when making categorization decisions. This article used computational modeling of behavioral data to take a closer look at this possibility in the context of a categorization task with redundancy. Iterative decision bomid modeling and drift diffusion models were used to detect individual differences in human categorization performance. The results show that participants differ in terms of what stimulus features they learned and how they use the learned features. For example, while some participants only learn one stimulus dimension (which is sufficient for perfect accuracy), others learn both stimulus dimensions (which is not required for perfect accuracy). Among participants that learned both dimensions, some used both dimensions, while others show error and RT patterns suggesting the use of only one of the dimensions. The diversity of obtained results is problematic for existing categorization models and suggests that each categorization model may be able to account for the performance of some but not all participants.