Modeling awake hippocampal reactivations with model-based bidirectional search

An Improved Dyna-Q Algorithm Inspired by the Forward Prediction Mechanism in the Rat Brain for Mobile Robot Path Planning

The traditional Model-Based Reinforcement Learning (MBRL) algorithm has high computational cost, poor convergence, and poor performance in robot spatial cognition and navigation tasks, and it cannot fully explain the ability of animals to quickly adapt to environmental changes and learn a variety of complex tasks. Studies have shown that vicarious trial and error (VTE) and the hippocampus forward prediction mechanism in rats and other mammals can be used as key components of action selection in MBRL to support "goal-oriented" behavior. Therefore, we propose an improved Dyna-Q algorithm inspired by the forward prediction mechanism of the hippocampus to solve the above problems and tackle the exploration-exploitation dilemma of Reinforcement Learning (RL). This algorithm alternately presents the potential path in the future for mobile robots and dynamically adjusts the sweep length according to the decision certainty, so as to determine action selection. We test the performance of the algorithm in a two-dimensional maze environment with static and dynamic obstacles, respectively. Compared with classic RL algorithms like State-Action-Reward-State-Action (SARSA) and Dyna-Q, the algorithm can speed up spatial cognition and improve the global search ability of path planning. In addition, our method reflects key features of how the brain organizes MBRL to effectively solve difficult tasks such as navigation, and it provides a new idea for spatial cognitive tasks from a biological perspective.

Model-Based and Model-Free Replay Mechanisms for Reinforcement Learning in Neurorobotics

Experience replay is widely used in AI to bootstrap reinforcement learning (RL) by enabling an agent to remember and reuse past experiences. Classical techniques include shuffled-, reversed-ordered- and prioritized-memory buffers, which have different properties and advantages depending on the nature of the data and problem. Interestingly, recent computational neuroscience work has shown that these techniques are relevant to model hippocampal reactivations recorded during rodent navigation. Nevertheless, the brain mechanisms for orchestrating hippocampal replay are still unclear. In this paper, we present recent neurorobotics research aiming to endow a navigating robot with a neuro-inspired RL architecture (including different learning strategies, such as model-based (MB) and model-free (MF), and different replay techniques). We illustrate through a series of numerical simulations how the specificities of robotic experimentation (e.g., autonomous state decomposition by the robot, noisy perception, state transition uncertainty, non-stationarity) can shed new lights on which replay techniques turn out to be more efficient in different situations. Finally, we close the loop by raising new hypotheses for neuroscience from such robotic models of hippocampal replay.

Replay in minds and machines

Experience-related brain activity patterns reactivate during sleep, wakeful rest, and brief pauses from active behavior. In parallel, machine learning research has found that experience replay can lead to substantial performance improvements in artificial agents. Together, these lines of research suggest replay has a variety of computational benefits for decision-making and learning. Here, we provide an overview of putative computational functions of replay as suggested by machine learning and neuroscientific research. We show that replay can lead to faster learning, less forgetting, reorganization or augmentation of experiences, and support planning and generalization. In addition, we highlight the benefits of reactivating abstracted internal representations rather than veridical memories, and discuss how replay could provide a mechanism to build internal representations that improve learning and decision-making.

Reward prediction errors drive declarative learning irrespective of agency

Recent years have witnessed a steady increase in the number of studies investigating the role of reward prediction errors (RPEs) in declarative learning. Specifically, in several experimental paradigms, RPEs drive declarative learning, with larger and more positive RPEs enhancing declarative learning. However, it is unknown whether this RPE must derive from the participant's own response, or whether instead, any RPE is sufficient to obtain the learning effect. To test this, we generated RPEs in the same experimental paradigm where we combined an agency and a nonagency condition. We observed no interaction between RPE and agency, suggesting that any RPE (irrespective of its source) can drive declarative learning. This result holds implications for declarative learning theory.

From spatial navigation via visual construction to episodic memory and imagination

This hybrid of review and personal essay argues that models of visual construction are essential to extend spatial navigation models to models that link episodic memory and imagination. The starting point is the TAM-WG model, combining the Taxon Affordance Model and the World Graph model of spatial navigation. The key here is to reject approaches in which memory is restricted to unanalyzed views from familiar places, and their later recall. Instead, we will seek mechanisms for imagining truly novel scenes and episodes. We thus introduce a specific variant of schema theory and VISIONS, a cooperative computation model of visual scene understanding in which a scene is represented by an assemblage of schema instances with links to lower-level "patches" of relevant visual data. We sketch a new conceptual framework for future modeling, Visual Integration of Diverse Multi-Modal Aspects, by extending VISIONS from static scenes to episodes combining agents, actions and objects and assess its relevance to both navigation and episodic memory. We can then analyze imagination as a constructive process that combines aspects of memories of prior episodes along with other schemas and adjusts them into a coherent whole which, through expectations associated with diverse episodes and schemas, may yield the linkage of episodes that constitutes a dream or a narrative. The result is IBSEN, a conceptual model of Imagination in Brain Systems for Episodes and Navigation. The essay closes by analyzing other papers in this Special Issue to assess to what extent their results relate to the research proposed here.