Learning and forgetting using reinforced Bayesian change detection

Agents living in volatile environments must be able to detect changes in contingencies while refraining to adapt to unexpected events that are caused by noise. In Reinforcement Learning (RL) frameworks, this requires learning rates that adapt to past reliability of the model. The observation that behavioural flexibility in animals tends to decrease following prolonged training in stable environment provides experimental evidence for such adaptive learning rates. However, in classical RL models, learning rate is either fixed or scheduled and can thus not adapt dynamically to environmental changes. Here, we propose a new Bayesian learning model, using variational inference, that achieves adaptive change detection by the use of Stabilized Forgetting, updating its current belief based on a mixture of fixed, initial priors and previous posterior beliefs. The weight given to these two sources is optimized alongside the other parameters, allowing the model to adapt dynamically to changes in environmental volatility and to unexpected observations. This approach is used to implement the “critic” of an actor-critic RL model, while the actor samples the resulting value distributions to choose which action to undertake. We show that our model can emulate different adaptation strategies to contingency changes, depending on its prior assumptions of environmental stability, and that model parameters can be fit to real data with high accuracy. The model also exhibits trade-offs between flexibility and computational costs that mirror those observed in real data. Overall, the proposed method provides a general framework to study learning flexibility and decision making in RL contexts.Author summary: In stable contexts, animals and humans exhibit automatic behaviour that allows them to make fast decisions. However, these automatic processes exhibit a lack of flexibility when environmental contingencies change. In the present paper, we propose a model of behavioural automatization that is based on adaptive forgetting and that emulates these properties. The model builds an estimate of the stability of the environment and uses this estimate to adjust its learning rate and the balance between exploration and exploitation policies. The model performs Bayesian inference on latent variables that represent relevant environmental properties, such as reward functions, optimal policies or environment stability. From there, the model makes decisions in order to maximize long-term rewards, with a noise proportional to environmental uncertainty. This rich model encompasses many aspects of Reinforcement Learning (RL), such as Temporal Difference RL and counterfactual learning, and accounts for the reduced computational cost of automatic behaviour. Using simulations, we show that this model leads to interesting predictions about the efficiency with which subjects adapt to sudden change of contingencies after prolonged training.