Learning to Perceive the World as Probabilistic or Deterministic via Interaction With Others: A Neuro-Robotics Experiment

An Overview of the Free Energy Principle and Related Research

The free energy principle and its corollary, the active inference framework, serve as theoretical foundations in the domain of neuroscience, explaining the genesis of intelligent behavior. This principle states that the processes of perception, learning, and decision making-within an agent-are all driven by the objective of "minimizing free energy," evincing the following behaviors: learning and employing a generative model of the environment to interpret observations, thereby achieving perception, and selecting actions to maintain a stable preferred state and minimize the uncertainty about the environment, thereby achieving decision making. This fundamental principle can be used to explain how the brain processes perceptual information, learns about the environment, and selects actions. Two pivotal tenets are that the agent employs a generative model for perception and planning and that interaction with the world (and other agents) enhances the performance of the generative model and augments perception. With the evolution of control theory and deep learning tools, agents based on the FEP have been instantiated in various ways across different domains, guiding the design of a multitude of generative models and decision-making algorithms. This letter first introduces the basic concepts of the FEP, followed by its historical development and connections with other theories of intelligence, and then delves into the specific application of the FEP to perception and decision making, encompassing both low-dimensional simple situations and high-dimensional complex situations. It compares the FEP with model-based reinforcement learning to show that the FEP provides a better objective function. We illustrate this using numerical studies of Dreamer3 by adding expected information gain into the standard objective function. In a complementary fashion, existing reinforcement learning, and deep learning algorithms can also help implement the FEP-based agents. Finally, we discuss the various capabilities that agents need to possess in complex environments and state that the FEP can aid agents in acquiring these capabilities.

A decoupled Bayesian method for snake robot control in unstructured environment

This paper presents a method which avoids the common practice of using a complex coupled snake robot model and performing kinematic analysis for control in cluttered environments. Instead, we introduce a completely decoupled dynamical Bayesian formulation with respect to interacted snake robot links and environmental objects, which requires much lower complexity for efficient and robust control. When a snake robot does not interact with obstacles, it runs by a simple serpenoid controller. However, when it exhibits interaction with environments, defined as close proximity or collision with targets and/or obstacles, we extend the conventional Bayesian framework by modeling such interactions in terms of stimuli. The proposed 'multi-neural-stimulus function' represents the cumulative effect of both external environmental influences and internal constraints of the snake robot. It implicitly handles the 'unexpected collision' problem and thus solve the difficult data association and shape adjustment problems for snake robot control in an innovative way. Preliminary experimental results have demonstrated promising performance of the proposed method comparing with the state-of-the-art.

Simulating developmental diversity: Impact of neural stochasticity on atypical flexibility and hierarchy

Introduction

Investigating the pathological mechanisms of developmental disorders is a challenge because the symptoms are a result of complex and dynamic factors such as neural networks, cognitive behavior, environment, and developmental learning. Recently, computational methods have started to provide a unified framework for understanding developmental disorders, enabling us to describe the interactions among those multiple factors underlying symptoms. However, this approach is still limited because most studies to date have focused on cross-sectional task performance and lacked the perspectives of developmental learning. Here, we proposed a new research method for understanding the mechanisms of the acquisition and its failures in hierarchical Bayesian representations using a state-of-the-art computational model, referred to as in silico neurodevelopment framework for atypical representation learning.

Methods

Simple simulation experiments were conducted using the proposed framework to examine whether manipulating the neural stochasticity and noise levels in external environments during the learning process can lead to the altered acquisition of hierarchical Bayesian representation and reduced flexibility.

Results

Networks with normal neural stochasticity acquired hierarchical representations that reflected the underlying probabilistic structures in the environment, including higher-order representation, and exhibited good behavioral and cognitive flexibility. When the neural stochasticity was high during learning, top-down generation using higher-order representation became atypical, although the flexibility did not differ from that of the normal stochasticity settings. However, when the neural stochasticity was low in the learning process, the networks demonstrated reduced flexibility and altered hierarchical representation. Notably, this altered acquisition of higher-order representation and flexibility was ameliorated by increasing the level of noises in external stimuli.

Discussion

These results demonstrated that the proposed method assists in modeling developmental disorders by bridging between multiple factors, such as the inherent characteristics of neural dynamics, acquisitions of hierarchical representation, flexible behavior, and external environment.

Learning from humans to build social cognition among robots

Self-organized groups of robots have generally coordinated their behaviors using quite simple social interactions. Although simple interactions are sufficient for some group behaviors, future research needs to investigate more elaborate forms of coordination, such as social cognition, to progress towards real deployments. In this perspective, we define social cognition among robots as the combination of social inference, social learning, social influence, and knowledge transfer, and propose that these abilities can be established in robots by building underlying mechanisms based on behaviors observed in humans. We review key social processes observed in humans that could inspire valuable capabilities in robots and propose that relevant insights from human social cognition can be obtained by studying human-controlled avatars in virtual environments that have the correct balance of embodiment and constraints. Such environments need to allow participants to engage in embodied social behaviors, for instance through situatedness and bodily involvement, but, at the same time, need to artificially constrain humans to the operational conditions of robots, for instance in terms of perception and communication. We illustrate our proposed experimental method with example setups in a multi-user virtual environment.

Interaction between Functional Connectivity and Neural Excitability in Autism: A Novel Framework for Computational Modeling and Application to Biological Data

Functional connectivity (FC) and neural excitability may interact to affect symptoms of autism spectrum disorder (ASD). We tested this hypothesis with neural network simulations, and applied it with functional magnetic resonance imaging (fMRI). A hierarchical recurrent neural network embodying predictive processing theory was subjected to a facial emotion recognition task. Neural network simulations examined the effects of FC and neural excitability on changes in neural representations by developmental learning, and eventually on ASD-like performance. Next, by mapping each neural network condition to subject subgroups on the basis of fMRI parameters, the association between ASD-like performance in the simulation and ASD diagnosis in the corresponding subject subgroup was examined. In the neural network simulation, the more homogeneous the neural excitability of the lower-level network, the more ASD-like the performance (reduced generalization and emotion recognition capability). In addition, in homogeneous networks, the higher the FC, the more ASD-like performance, while in heterogeneous networks, the higher the FC, the less ASD-like performance, demonstrating that FC and neural excitability interact. As an underlying mechanism, neural excitability determines the generalization capability of top-down prediction, and FC determines whether the model's information processing will be top-down prediction-dependent or bottom-up sensory-input dependent. In fMRI datasets, ASD was actually more prevalent in subject subgroups corresponding to the network condition showing ASD-like performance. The current study suggests an interaction between FC and neural excitability, and presents a novel framework for computational modeling and biological application of a developmental learning process underlying cognitive alterations in ASD.

Leveraging Motor Babbling for Efficient Robot Learning

Deep robotic learning by learning from demonstration allows robots to mimic a given demonstration and generalize their performance to unknown task setups. However, this generalization ability is heavily affected by the number of demonstrations, which can be costly to manually generate. Without sufficient demonstrations, robots tend to overfit to the available demonstrations and lose the robustness offered by deep learning. Applying the concept of motor babbling – a process similar to that by which human infants move their bodies randomly to obtain proprioception – is also effective for allowing robots to enhance their generalization ability. Furthermore, the generation of babbling data is simpler than task-oriented demonstrations. Previous researches use motor babbling in the concept of pre-training and fine-tuning but have the problem of the babbling data being overwritten by the task data. In this work, we propose an RNN-based robot-control framework capable of leveraging targetless babbling data to aid the robot in acquiring proprioception and increasing the generalization ability of the learned task data by learning both babbling and task data simultaneously. Through simultaneous learning, our framework can use the dynamics obtained from babbling data to learn the target task efficiently. In the experiment, we prepare demonstrations of a block-picking task and aimless-babbling data. With our framework, the robot can learn tasks faster and show greater generalization ability when blocks are at unknown positions or move during execution.

Emergence of Content-Agnostic Information Processing by a Robot Using Active Inference, Visual Attention, Working Memory, and Planning

Generalization by learning is an essential cognitive competency for humans. For example, we can manipulate even unfamiliar objects and can generate mental images before enacting a preplan. How is this possible? Our study investigated this problem by revisiting our previous study (Jung, Matsumoto, & Tani, 2019), which examined the problem of vision-based, goal-directed planning by robots performing a task of block stacking. By extending the previous study, our work introduces a large network comprising dynamically interacting submodules, including visual working memory (VWMs), a visual attention module, and an executive network. The executive network predicts motor signals, visual images, and various controls for attention, as well as masking of visual information. The most significant difference from the previous study is that our current model contains an additional VWM. The entire network is trained by using predictive coding and an optimal visuomotor plan to achieve a given goal state is inferred using active inference. Results indicate that our current model performs significantly better than that used in Jung et al. (2019), especially when manipulating blocks with unlearned colors and textures. Simulation results revealed that the observed generalization was achieved because content-agnostic information processing developed through synergistic interaction between the second VWM and other modules during the course of learning, in which memorizing image contents and transforming them are dissociated. This letter verifies this claim by conducting both qualitative and quantitative analysis of simulation results.

Neural network modeling of altered facial expression recognition in autism spectrum disorders based on predictive processing framework

The mechanism underlying the emergence of emotional categories from visual facial expression information during the developmental process is largely unknown. Therefore, this study proposes a system-level explanation for understanding the facial emotion recognition process and its alteration in autism spectrum disorder (ASD) from the perspective of predictive processing theory. Predictive processing for facial emotion recognition was implemented as a hierarchical recurrent neural network (RNN). The RNNs were trained to predict the dynamic changes of facial expression movies for six basic emotions without explicit emotion labels as a developmental learning process, and were evaluated by the performance of recognizing unseen facial expressions for the test phase. In addition, the causal relationship between the network characteristics assumed in ASD and ASD-like cognition was investigated. After the developmental learning process, emotional clusters emerged in the natural course of self-organization in higher-level neurons, even though emotional labels were not explicitly instructed. In addition, the network successfully recognized unseen test facial sequences by adjusting higher-level activity through the process of minimizing precision-weighted prediction error. In contrast, the network simulating altered intrinsic neural excitability demonstrated reduced generalization capability and impaired emotional clustering in higher-level neurons. Consistent with previous findings from human behavioral studies, an excessive precision estimation of noisy details underlies this ASD-like cognition. These results support the idea that impaired facial emotion recognition in ASD can be explained by altered predictive processing, and provide possible insight for investigating the neurophysiological basis of affective contact.

Hypersphere-Based Weight Imprinting for Few-Shot Learning on Embedded Devices

Weight imprinting (WI) was recently introduced as a way to perform gradient descent-free few-shot learning. Due to this, WI was almost immediately adapted for performing few-shot learning on embedded neural network accelerators that do not support back-propagation, e.g., edge tensor processing units. However, WI suffers from many limitations, e.g., it cannot handle novel categories with multimodal distributions and special care should be given to avoid overfitting the learned embeddings on the training classes since this can have a devastating effect on classification accuracy (for the novel categories). In this article, we propose a novel hypersphere-based WI approach that is capable of training neural networks in a regularized, imprinting-aware way effectively overcoming the aforementioned limitations. The effectiveness of the proposed method is demonstrated using extensive experiments on three image data sets.

Neurobiologically Inspired Self-Monitoring Systems

In this article, we explore neurobiological principles that could be deployed in systems requiring self-preservation, adaptive control, and contextual awareness. We start with low-level control for sensor processing and motor reflexes. We then discuss how critical it is at an intermediate level to maintain homeostasis and predict system set points. We end with a discussion at a high-level, or cognitive level, where planning and prediction can further monitor the system and optimize performance. We emphasize the information flow between these levels both from a systems neuroscience and an engineering point of view. Throughout the paper, we describe the brain systems that carry out these functions and provide examples from artificial intelligence, machine learning, and robotics that include these features. Our goal is to show how biological organisms performing self-monitoring can inspire the design of autonomous and embedded systems.

A Novel Predictive-Coding-Inspired Variational RNN Model for Online Prediction and Recognition

This study introduces PV-RNN, a novel variational RNN inspired by predictive-coding ideas. The model learns to extract the probabilistic structures hidden in fluctuating temporal patterns by dynamically changing the stochasticity of its latent states. Its architecture attempts to address two major concerns of variational Bayes RNNs: how latent variables can learn meaningful representations and how the inference model can transfer future observations to the latent variables. PV-RNN does both by introducing adaptive vectors mirroring the training data, whose values can then be adapted differently during evaluation. Moreover, prediction errors during backpropagation-rather than external inputs during the forward computation-are used to convey information to the network about the external data. For testing, we introduce error regression for predicting unseen sequences as inspired by predictive coding that leverages those mechanisms. As in other variational Bayes RNNs, our model learns by maximizing a lower bound on the marginal likelihood of the sequential data, which is composed of two terms: the negative of the expectation of prediction errors and the negative of the Kullback-Leibler divergence between the prior and the approximate posterior distributions. The model introduces a weighting parameter, the meta-prior, to balance the optimization pressure placed on those two terms. We test the model on two data sets with probabilistic structures and show that with high values of the meta-prior, the network develops deterministic chaos through which the randomness of the data is imitated. For low values, the model behaves as a random process. The network performs best on intermediate values and is able to capture the latent probabilistic structure with good generalization. Analyzing the meta-prior's impact on the network allows us to precisely study the theoretical value and practical benefits of incorporating stochastic dynamics in our model. We demonstrate better prediction performance on a robot imitation task with our model using error regression compared to a standard variational Bayes model lacking such a procedure.

Neurorobotics-A Thriving Community and a Promising Pathway Toward Intelligent Cognitive Robots

Neurorobots are robots whose control has been modeled after some aspect of the brain. Since the brain is so closely coupled to the body and situated in the environment, Neurorobots can be a powerful tool for studying neural function in a holistic fashion. It may also be a means to develop autonomous systems that have some level of biological intelligence. The present article provides my perspective on this field, points out some of the landmark events, and discusses its future potential.

Predictive Coding for Dynamic Visual Processing: Development of Functional Hierarchy in a Multiple Spatiotemporal Scales RNN Model

This letter proposes a novel predictive coding type neural network model, the predictive multiple spatiotemporal scales recurrent neural network (P-MSTRNN). The P-MSTRNN learns to predict visually perceived human whole-body cyclic movement patterns by exploiting multiscale spatiotemporal constraints imposed on network dynamics by using differently sized receptive fields as well as different time constant values for each layer. After learning, the network can imitate target movement patterns by inferring or recognizing corresponding intentions by means of the regression of prediction error. Results show that the network can develop a functional hierarchy by developing a different type of dynamic structure at each layer. The letter examines how model performance during pattern generation, as well as predictive imitation, varies depending on the stage of learning. The number of limit cycle attractors corresponding to target movement patterns increases as learning proceeds. Transient dynamics developing early in the learning process successfully perform pattern generation and predictive imitation tasks. The letter concludes that exploitation of transient dynamics facilitates successful task performance during early learning periods.

Dynamical Integration of Language and Behavior in a Recurrent Neural Network for Human-Robot Interaction

To work cooperatively with humans by using language, robots must not only acquire a mapping between language and their behavior but also autonomously utilize the mapping in appropriate contexts of interactive tasks online. To this end, we propose a novel learning method linking language to robot behavior by means of a recurrent neural network. In this method, the network learns from correct examples of the imposed task that are given not as explicitly separated sets of language and behavior but as sequential data constructed from the actual temporal flow of the task. By doing this, the internal dynamics of the network models both language-behavior relationships and the temporal patterns of interaction. Here, "internal dynamics" refers to the time development of the system defined on the fixed-dimensional space of the internal states of the context layer. Thus, in the execution phase, by constantly representing where in the interaction context it is as its current state, the network autonomously switches between recognition and generation phases without any explicit signs and utilizes the acquired mapping in appropriate contexts. To evaluate our method, we conducted an experiment in which a robot generates appropriate behavior responding to a human's linguistic instruction. After learning, the network actually formed the attractor structure representing both language-behavior relationships and the task's temporal pattern in its internal dynamics. In the dynamics, language-behavior mapping was achieved by the branching structure. Repetition of human's instruction and robot's behavioral response was represented as the cyclic structure, and besides, waiting to a subsequent instruction was represented as the fixed-point attractor. Thanks to this structure, the robot was able to interact online with a human concerning the given task by autonomously switching phases.