Predictive representations can link model-based reinforcement learning to model-free mechanisms.
Ontology highlight
ABSTRACT:
SUBMITTER: Russek EM
PROVIDER: S-EPMC5628940 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:In computer science, reinforcement learning is a powerful framework with which artificial agents can learn to maximize their performance for any given Markov decision process (MDP). Advances over the last decade, in combination with deep neural networks, have enjoyed performance advantages over humans in many difficult task settings. However, such frameworks perform far less favorably when evaluated in their ability to generalize or transfer representations across different tasks. Existing algorithms that facilitate transfer typically are limited to cases in which the transition function or the optimal policy is portable to new contexts, but achieving "deep transfer" characteristic of human behavior has been elusive. Such transfer typically requires discovery of abstractions that permit analogical reuse of previously learned representations to superficially distinct tasks. Here, we demonstrate that abstractions that minimize error in predictions of reward outcomes generalize across tasks with different transition and reward functions. Such reward-predictive representations compress the state space of a task into a lower dimensional representation by combining states that are equivalent in terms of both the transition and reward functions. Because only state equivalences are considered, the resulting state representation is not tied to the transition and reward functions themselves and thus generalizes across tasks with different reward and transition functions. These results contrast with those using abstractions that myopically maximize reward in any given MDP and motivate further experiments in humans and animals to investigate if neural and cognitive systems involved in state representation perform abstractions that facilitate such equivalence relations.
| S-EPMC7591094 | biostudies-literature
Project description:Considerable evidence suggests that multiple learning systems can drive behavior. Choice can proceed reflexively from previous actions and their associated outcomes, as captured by "model-free" learning algorithms, or flexibly from prospective consideration of outcomes that might occur, as captured by "model-based" learning algorithms. However, differential contributions of dopamine to these systems are poorly understood. Dopamine is widely thought to support model-free learning by modulating plasticity in striatum. Model-based learning may also be affected by these striatal effects, or by other dopaminergic effects elsewhere, notably on prefrontal working memory function. Indeed, prominent demonstrations linking striatal dopamine to putatively model-free learning did not rule out model-based effects, whereas other studies have reported dopaminergic modulation of verifiably model-based learning, but without distinguishing a prefrontal versus striatal locus. To clarify the relationships between dopamine, neural systems, and learning strategies, we combine a genetic association approach in humans with two well-studied reinforcement learning tasks: one isolating model-based from model-free behavior and the other sensitive to key aspects of striatal plasticity. Prefrontal function was indexed by a polymorphism in the COMT gene, differences of which reflect dopamine levels in the prefrontal cortex. This polymorphism has been associated with differences in prefrontal activity and working memory. Striatal function was indexed by a gene coding for DARPP-32, which is densely expressed in the striatum where it is necessary for synaptic plasticity. We found evidence for our hypothesis that variations in prefrontal dopamine relate to model-based learning, whereas variations in striatal dopamine function relate to model-free learning.Decisions can stem reflexively from their previously associated outcomes or flexibly from deliberative consideration of potential choice outcomes. Research implicates a dopamine-dependent striatal learning mechanism in the former type of choice. Although recent work has indicated that dopamine is also involved in flexible, goal-directed decision-making, it remains unclear whether it also contributes via striatum or via the dopamine-dependent working memory function of prefrontal cortex. We examined genetic indices of dopamine function in these regions and their relation to the two choice strategies. We found that striatal dopamine function related most clearly to the reflexive strategy, as previously shown, and that prefrontal dopamine related most clearly to the flexible strategy. These findings suggest that dissociable brain regions support dissociable choice strategies.
| S-EPMC4728725 | biostudies-literature
Project description:The Wisconsin Card Sorting Test (WCST) is considered a gold standard for the assessment of cognitive flexibility. On the WCST, repeating a sorting category following negative feedback is typically treated as indicating reduced cognitive flexibility. Therefore such responses are referred to as 'perseveration' errors. Recent research suggests that the propensity for perseveration errors is modulated by response demands: They occur less frequently when their commitment repeats the previously executed response. Here, we propose parallel reinforcement-learning models of card sorting performance, which assume that card sorting performance can be conceptualized as resulting from model-free reinforcement learning at the level of responses that occurs in parallel with model-based reinforcement learning at the categorical level. We compared parallel reinforcement-learning models with purely model-based reinforcement learning, and with the state-of-the-art attentional-updating model. We analyzed data from 375 participants who completed a computerized WCST. Parallel reinforcement-learning models showed best predictive accuracies for the majority of participants. Only parallel reinforcement-learning models accounted for the modulation of perseveration propensity by response demands. In conclusion, parallel reinforcement-learning models provide a new theoretical perspective on card sorting and it offers a suitable framework for discerning individual differences in latent processes that subserve behavioral flexibility.
| S-EPMC7508815 | biostudies-literature
Project description:From both functional and biological considerations, it is widely believed that action production, planning, and goal-oriented behaviors supported by the frontal cortex are organized hierarchically [Fuster (1991); Koechlin, E., Ody, C., & Kouneiher, F. (2003). Neuroscience: The architecture of cognitive control in the human prefrontal cortex. Science, 424, 1181-1184; Miller, G. A., Galanter, E., & Pribram, K. H. (1960). Plans and the structure of behavior. New York: Holt]. However, the nature of the different levels of the hierarchy remains unclear, and little attention has been paid to the origins of such a hierarchy. We address these issues through biologically-inspired computational models that develop representations through reinforcement learning. We explore several different factors in these models that might plausibly give rise to a hierarchical organization of representations within the PFC, including an initial connectivity hierarchy within PFC, a hierarchical set of connections between PFC and subcortical structures controlling it, and differential synaptic plasticity schedules. Simulation results indicate that architectural constraints contribute to the segregation of different types of representations, and that this segregation facilitates learning. These findings are consistent with the idea that there is a functional hierarchy in PFC, as captured in our earlier computational models of PFC function and a growing body of empirical data.
| S-EPMC2783795 | biostudies-literature
Project description:Organisms appear to learn and make decisions using different strategies known as model-free and model-based learning; the former is mere reinforcement of previously rewarded actions and the latter is a forward-looking strategy that involves evaluation of action-state transition probabilities. Prior work has used neural data to argue that both model-based and model-free learners implement a value comparison process at trial onset, but model-based learners assign more weight to forward-looking computations. Here using eye-tracking, we report evidence for a different interpretation of prior results: model-based subjects make their choices prior to trial onset. In contrast, model-free subjects tend to ignore model-based aspects of the task and instead seem to treat the decision problem as a simple comparison process between two differentially valued items, consistent with previous work on sequential-sampling models of decision making. These findings illustrate a problem with assuming that experimental subjects make their decisions at the same prescribed time.
| S-EPMC4987535 | biostudies-literature
Project description:Reinforcement learning (RL) uses sequential experience with situations ("states") and outcomes to assess actions. Whereas model-free RL uses this experience directly, in the form of a reward prediction error (RPE), model-based RL uses it indirectly, building a model of the state transition and outcome structure of the environment, and evaluating actions by searching this model. A state prediction error (SPE) plays a central role, reporting discrepancies between the current model and the observed state transitions. Using functional magnetic resonance imaging in humans solving a probabilistic Markov decision task, we found the neural signature of an SPE in the intraparietal sulcus and lateral prefrontal cortex, in addition to the previously well-characterized RPE in the ventral striatum. This finding supports the existence of two unique forms of learning signal in humans, which may form the basis of distinct computational strategies for guiding behavior.
| S-EPMC2895323 | biostudies-other
Project description:Temporal-difference (TD) algorithms have been proposed as models of reinforcement learning (RL). We examine two issues of distributed representation in these TD algorithms: distributed representations of belief and distributed discounting factors. Distributed representation of belief allows the believed state of the world to distribute across sets of equivalent states. Distributed exponential discounting factors produce hyperbolic discounting in the behavior of the agent itself. We examine these issues in the context of a TD RL model in which state-belief is distributed over a set of exponentially-discounting "micro-Agents", each of which has a separate discounting factor (gamma). Each microAgent maintains an independent hypothesis about the state of the world, and a separate value-estimate of taking actions within that hypothesized state. The overall agent thus instantiates a flexible representation of an evolving world-state. As with other TD models, the value-error (delta) signal within the model matches dopamine signals recorded from animals in standard conditioning reward-paradigms. The distributed representation of belief provides an explanation for the decrease in dopamine at the conditioned stimulus seen in overtrained animals, for the differences between trace and delay conditioning, and for transient bursts of dopamine seen at movement initiation. Because each microAgent also includes its own exponential discounting factor, the overall agent shows hyperbolic discounting, consistent with behavioral experiments.
| S-EPMC2760757 | biostudies-literature
Project description:IntroductionObsessive-compulsive disorder (OCD) is characterized by an imbalance between goal-directed and habitual learning systems in behavioral control, but it is unclear whether these impairments are due to a single system abnormality of the goal-directed system or due to an impairment in a separate arbitration mechanism that selects which system controls behavior at each point in time.MethodsA total of 30 OCD patients and 120 healthy controls performed a 2-choice, 3-stage Markov decision-making paradigm. Reinforcement learning models were used to estimate goal-directed learning (as model-based reinforcement learning) and habitual learning (as model-free reinforcement learning). In general, 29 high Obsessive-Compulsive Inventory-Revised (OCI-R) score controls, 31 low OCI-R score controls, and all 30 OCD patients were selected for the analysis.ResultsObsessive-compulsive disorder (OCD) patients showed less appropriate strategy choices than controls regardless of whether the OCI-R scores in the control subjects were high (p = 0.012) or low (p < 0.001), specifically showing a greater model-free strategy use in task conditions where the model-based strategy was optimal. Furthermore, OCD patients (p = 0.001) and control subjects with high OCI-R scores (H-OCI-R; p = 0.009) both showed greater system switching rather than consistent strategy use in task conditions where model-free use was optimal.ConclusionThese findings indicated an impaired arbitration mechanism for flexible adaptation to environmental demands in both OCD patients and healthy individuals reporting high OCI-R scores.
| S-EPMC10250695 | biostudies-literature
Project description:To behave adaptively, animals must learn to predict future reward, or value. To do this, animals are thought to learn reward predictions using reinforcement learning. However, in contrast to classical models, animals must learn to estimate value using only incomplete state information. Previous work suggests that animals estimate value in partially observable tasks by first forming "beliefs"-optimal Bayesian estimates of the hidden states in the task. Although this is one way to solve the problem of partial observability, it is not the only way, nor is it the most computationally scalable solution in complex, real-world environments. Here we show that a recurrent neural network (RNN) can learn to estimate value directly from observations, generating reward prediction errors that resemble those observed experimentally, without any explicit objective of estimating beliefs. We integrate statistical, functional, and dynamical systems perspectives on beliefs to show that the RNN's learned representation encodes belief information, but only when the RNN's capacity is sufficiently large. These results illustrate how animals can estimate value in tasks without explicitly estimating beliefs, yielding a representation useful for systems with limited capacity.
| S-EPMC10513382 | biostudies-literature
Project description:Previous studies found that stress shifts behavioral control by promoting habits while decreasing goal-directed behaviors during reward-based decision-making. It is, however, unclear how stress disrupts the relative contribution of the two systems controlling reward-seeking behavior, i.e. model-free (or habit) and model-based (or goal-directed). Here, we investigated whether stress biases the contribution of model-free and model-based reinforcement learning processes differently depending on the valence of outcome, and whether stress alters the learning rate, i.e., how quickly information from the new environment is incorporated into choices. Participants were randomly assigned to either a stress or a control condition, and performed a two-stage Markov decision-making task in which the reward probabilities underwent periodic reversals without notice. We found that stress increased the contribution of model-free reinforcement learning only after negative outcome. Furthermore, stress decreased the learning rate. The results suggest that stress diminishes one's ability to make adaptive choices in multiple aspects of reinforcement learning. This finding has implications for understanding how stress facilitates maladaptive habits, such as addictive behavior, and other dysfunctional behaviors associated with stress in clinical and educational contexts.
| S-EPMC5516979 | biostudies-literature