Autonomous learning of features for control: Experiments with embodied and situated agents.
Ontology highlight
ABSTRACT:
SUBMITTER: Milano N
PROVIDER: S-EPMC8049238 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:We demonstrate how the evolutionary training of embodied agents can be extended with a curriculum learning algorithm that automatically selects the environmental conditions in which the evolving agents are evaluated. The environmental conditions are selected to adjust the level of difficulty to the ability level of the current evolving agents, and to challenge the weaknesses of the evolving agents. The method does not require domain knowledge and does not introduce additional hyperparameters. The results collected on two benchmark problems, that require to solve a task in significantly varying environmental conditions, demonstrate that the method proposed outperforms conventional learning methods and generates solutions which are robust to variations and able to cope with different environmental conditions.
| S-EPMC8076209 | biostudies-literature
Project description:Massive Open Online Courses are a dominant force in remote-learning yet suffer from persisting problems stemming from lack of commitment and low completion rates. In this initial study we investigate how the use of immersive virtual environments for Power-Point based informational learning may benefit learners and mimic traditional lectures successfully. We examine the role of embodied agent tutors which are frequently implemented within virtual learning environments. We find similar performance on a bespoke knowledge test and metrics for motivation, satisfaction, and engagement by learners in both real and virtual environments, regardless of embodied agent tutor presence. Our results raise questions regarding the viability of using virtual environments for remote-learning paradigms, and we emphasise the need for further investigation to inform the design of effective remote-learning applications.
| S-EPMC7924662 | biostudies-literature
Project description:From the perspective of constructivist theories, emotion results from learning assemblies of relevant perceptual, cognitive, interoceptive, and motor processes in specific situations. Across emotional experiences over time, learned assemblies of processes accumulate in memory that later underlie emotional experiences in similar situations. A neuroimaging experiment guided participants to experience (and thus learn) situated forms of emotion, and then assessed whether participants tended to experience situated forms of the emotion later. During the initial learning phase, some participants immersed themselves in vividly imagined fear and anger experiences involving physical harm, whereas other participants immersed themselves in vividly imagined fear and anger experiences involving negative social evaluation. In the subsequent testing phase, both learning groups experienced fear and anger while their neural activity was assessed with functional magnetic resonance imaging (fMRI). A variety of results indicated that the physical and social learning groups incidentally learned different situated forms of a given emotion. Consistent with constructivist theories, these findings suggest that learning plays a central role in emotion, with emotion adapted to the situations in which it is experienced.
| S-EPMC6037607 | biostudies-literature
Project description:Coordination in groups faces a sub-optimization problem and theory suggests that some randomness may help to achieve global optima. Here we performed experiments involving a networked colour coordination game in which groups of humans interacted with autonomous software agents (known as bots). Subjects (n?=?4,000) were embedded in networks (n?=?230) of 20 nodes, to which we sometimes added 3 bots. The bots were programmed with varying levels of behavioural randomness and different geodesic locations. We show that bots acting with small levels of random noise and placed in central locations meaningfully improve the collective performance of human groups, accelerating the median solution time by 55.6%. This is especially the case when the coordination problem is hard. Behavioural randomness worked not only by making the task of humans to whom the bots were connected easier, but also by affecting the gameplay of the humans among themselves and hence creating further cascades of benefit in global coordination in these heterogeneous systems.
| S-EPMC5912653 | biostudies-literature
Project description:Many plants autonomously change morphology and function in response to environmental stimuli or sequences of stimuli. In contrast with the electronically-integrated sensors, actuators, and microprocessors in traditional mechatronic systems, natural systems embody these sensing, actuation, and control functions within their compositional and structural features. Inspired by nature, we embody logic in autonomous systems to enable them to respond to multiple stimuli. Using 3D printable fibrous composites, we fabricate structures with geometries near bifurcation points associated with a transition between bistability and monostability. When suitable stimuli are present, the materials swell anisotropically. This forces a key geometric parameter to pass through a bifurcation, triggering rapid and large-amplitude self-actuation. The actuation time can be programmed by varying structural parameters (from 0.6 to 108 s for millimeter-scale structures). We demonstrate this bioinspired control strategy with examples that respond to their environment according to their embodied logic, without electronics, external control, or tethering.
| S-EPMC6328580 | biostudies-literature
Project description:We propose that the domain general process of categorization contributes to the perception of stress. When a situation contains features associated with stressful experiences, it is categorized as stressful. From the perspective of situated cognition, the features used to categorize experiences as stressful are the features typically true of stressful situations. To test this hypothesis, we asked participants to evaluate the perceived stress of 572 imagined situations, and to also evaluate each situation for how much it possessed 19 features potentially associated with stressful situations and their processing (e.g., self-threat, familiarity, visual imagery, outcome certainty). Following variable reduction through factor analysis, a core set of 8 features associated with stressful situations-expectation violation, self-threat, coping efficacy, bodily experience, arousal, negative valence, positive valence, and perseveration-all loaded on a single Core Stress Features factor. In a multilevel model, this factor and an Imagery factor explained 88% of the variance in judgments of perceived stress, with significant random effects reflecting differences in how individual participants categorized stress. These results support the hypothesis that people categorize situations as stressful to the extent that typical features of stressful situations are present. To our knowledge, this is the first attempt to establish a comprehensive set of features that predicts perceived stress.
| S-EPMC4987261 | biostudies-literature
Project description:Neural network simulation is an important tool for generating and evaluating hypotheses on the structure, dynamics, and function of neural circuits. For scientific questions addressing organisms operating autonomously in their environments, in particular where learning is involved, it is crucial to be able to operate such simulations in a closed-loop fashion. In such a set-up, the neural agent continuously receives sensory stimuli from the environment and provides motor signals that manipulate the environment or move the agent within it. So far, most studies requiring such functionality have been conducted with custom simulation scripts and manually implemented tasks. This makes it difficult for other researchers to reproduce and build upon previous work and nearly impossible to compare the performance of different learning architectures. In this work, we present a novel approach to solve this problem, connecting benchmark tools from the field of machine learning and state-of-the-art neural network simulators from computational neuroscience. The resulting toolchain enables researchers in both fields to make use of well-tested high-performance simulation software supporting biologically plausible neuron, synapse and network models and allows them to evaluate and compare their approach on the basis of standardized environments with various levels of complexity. We demonstrate the functionality of the toolchain by implementing a neuronal actor-critic architecture for reinforcement learning in the NEST simulator and successfully training it on two different environments from the OpenAI Gym. We compare its performance to a previously suggested neural network model of reinforcement learning in the basal ganglia and a generic Q-learning algorithm.
| S-EPMC6687756 | biostudies-literature
Project description:This work addresses the problem of reference tracking in autonomously learning robots with unknown, nonlinear dynamics. Existing solutions require model information or extensive parameter tuning, and have rarely been validated in real-world experiments. We propose a learning control scheme that learns to approximate the unknown dynamics by a Gaussian Process (GP), which is used to optimize and apply a feedforward control input on each trial. Unlike existing approaches, the proposed method neither requires knowledge of the system states and their dynamics nor knowledge of an effective feedback control structure. All algorithm parameters are chosen automatically, i.e. the learning method works plug and play. The proposed method is validated in extensive simulations and real-world experiments. In contrast to most existing work, we study learning dynamics for more than one motion task as well as the robustness of performance across a large range of learning parameters. The method's plug and play applicability is demonstrated by experiments with a balancing robot, in which the proposed method rapidly learns to track the desired output. Due to its model-agnostic and plug and play properties, the proposed method is expected to have high potential for application to a large class of reference tracking problems in systems with unknown, nonlinear dynamics.
| S-EPMC9315427 | biostudies-literature
Project description:We investigated how naturalistic actions in a highly immersive, multimodal, interactive 3D virtual reality (VR) environment may enhance word encoding by recording EEG in a pre/post-test learning paradigm. While behavior data have shown that coupling word encoding with gestures congruent with word meaning enhances learning, the neural underpinnings of this effect have yet to be elucidated. We coupled EEG recording with VR to examine whether embodied learning improves learning and creates linguistic representations that produce greater motor resonance. Participants learned action verbs in an L2 in two different conditions: specific action (observing and performing congruent actions on virtual objects) and pointing (observing actions and pointing to virtual objects). Pre- and post-training participants performed a match-mismatch task as we measured EEG (variation in the N400 response as a function of match between observed actions and auditory verbs) and a passive listening task while we measured motor activation (mu [8-13 Hz] and beta band [13-30 Hz] desynchronization during auditory verb processing) during verb processing. Contrary to our expectations, post-training results revealed neither semantic nor motor effects in either group when considered independently of learning success. Behavioral results showed a great deal of variability in learning success. When considering performance, low performance learners showed no semantic effect and high performance learners exhibited an N400 effect for mismatch versus match trials post-training, independent of the type of learning. Taken as a whole, our results suggest that embodied processes can play an important role in L2 learning.
| S-EPMC11192445 | biostudies-literature
Project description:In recent years, research on embodied cognition has inspired a number of studies on multimedia learning and instructional psychology. However, in contrast to traditional research on education and multimedia learning, studies on embodied learning (i.e., focusing on bodily action and perception in the context of education) in some cases pose new problems for the measurement of cognitive load. This review provides an overview over recent studies on embodied learning in which cognitive load was measured using surveys, behavioral data, or physiological measures. The different methods are assessed in terms of their success in finding differences of cognitive load in embodied learning scenarios. At the same time, we highlight the most important challenges for researchers aiming to include these measures into their study designs. The main issues we identified are: (1) Subjective measures must be appropriately phrased to be useful for embodied learning; (2) recent findings indicate potentials as well as problematic aspects of dual-task measures; (3) the use of physiological measures offers great potential, but may require mobile equipment in the context of embodied scenarios; (4) meta-cognitive measures can be useful extensions of cognitive load measurement for embodied learning.
| S-EPMC5539229 | biostudies-other