Rapid Learning Research Articles

A neural circuit architecture for rapid learning in goal-directed navigation

Anchoring goals to spatial representations enables flexible navigation but is challenging in novel environments when both representations must be acquired simultaneously. We propose a framework for how Drosophila uses internal representations of head direction (HD) to build goal representations upon selective thermal reinforcement. We show that flies use stochastically generated fixations and directed saccades to express heading preferences in an operant visual learning paradigm and that HD neurons are required to modify these preferences based on reinforcement. We used a symmetric visual setting to expose how flies' HD and goal representations co-evolve and how the reliability of these interacting representations impacts behavior. Finally, we describe how rapid learning of new goal headings may rest on a behavioral policy whose parameters are flexible but whose form is genetically encoded in circuit architecture. Such evolutionarily structured architectures, which enable rapidly adaptive behavior driven by internal representations, may be relevant across species.

Patient care in complex Sociotechnological ecosystems and learning health systems.

The learning health system (LHS) model was proposed to provide real-time, bi-directional flow of learning using data captured in health information technology systems to deliver rapid learning in healthcare delivery. As highlighted by the landmark National Academy of Medicine report "Crossing the Quality Chasm," the U.S. healthcare delivery industry represents complex adaptive systems, and there is an urgent need to develop innovative methods to identify efficient team structures by harnessing real-world care delivery data found in the electronic health record (EHR). We offer a discussion surrounding the complexities of team communication and how solutions may be guided by theories such as the Multiteam System (MTS) framework and the Multitheoretical Multilevel Framework of Communication Networks. To advance healthcare delivery science and promote LHSs, our team has been building a new line of research using EHR data to study MTS in the complex real world of cancer care delivery. We are developing new network metrics to study MTSs and will be analyzing the impact of EHR communication network structures on patient outcomes. As this research leads to patient care delivery interventions/tools, healthcare leaders and healthcare professionals can effectively use health IT data to implement the most evidence-based collaboration approaches in order to achieve the optimal LHS and patient outcomes.

A physics-inspired neural network for short-wave radiation parameterization

Abstract Radiation parameterization schemes are crucial components of weather and climate models, but they are also known to be computationally intensive. In recent decades, researchers have attempted to emulate these schemes using neural networks, with more attention to convolutional neural networks. However, in this paper, we explore the potential of recurrent neural networks (RNNs) for predicting solar heating rates. Our architecture was trained and tested using long-term hindcast data from the Pechora Sea region, with the conventional RRTMG scheme serving as a shortwave parameterization. Our findings show that the RNN offers rapid learning, fast inference, and excellent data fitting. We also present preliminary results demonstrating the use of RNNs for operational weather forecasting, which achieved a significant speedup in parameterization and reduced the overall forecast time by 40 %.

Point-of-care ultrasound (POCUS) pediatric resident training course: a cross-sectional survey

BackgroundPoint-of-care ultrasound (POCUS) is becoming increasingly crucial in the Pediatric Emergency Department for objective patient examination. However, despite its growing interest and wide-ranging applications, POCUS remains relatively unexplored in general pediatric training and education. Many physicians still find it challenging to comprehend and implement.MethodsA theoretical-practical POCUS course for pediatric residents was conducted at the University of Catania, Italy. The course’s effectiveness and practical impact on residents was assessed through a pre-post training survey. The first part of the questionnaire focused on the self-perceived time needed to learn how to recognize the following conditions using POCUS: (i) Pleural effusion (ii) Lung consolidation (iii) Pneumothorax (PNX) (iv) Cardiac contractility (v) Pericardial effusion (vi) Perisplenic effusion (vii) Morison’s pouch effusion (viii) Douglas’ pouch effusion (ix) Filling and collapsibility of the inferior vena cava. In the second part, we compared the potential role of POCUS in (i) Reducing the use of ionizing radiation in children (ii) Increasing the sense of security in diagnosis and treatment decisions making and (iii) Increasing the residents’ confidence level with POCUS after the course on a 1-to-10 rating scale.ResultsSeventy-two residents participated in the study. The statistical analysis showed significant pre-post differences in almost all the items considered, except for “cardiac contractility” and “PNX”. Furthermore, the perceived potential role of POCUS in reducing ionizing radiation usage and the sense of security in diagnosis and treatment decisions showed statistically significant differences (p < 0.05) before and after the course. Data analysis also revealed a consistently high confidence level with POCUS after the course.ConclusionsThe results highlight the importance of including a POCUS track course in pediatric post-graduate programs due to its simplicity, rapid learning time, and clinical usefulness. Based on these findings, it would be recommended to increase the teaching hours dedicated to the recognition of pneumothorax and cardiology POCUS examination. Emphasizing POCUS training in pediatric education can enhance patient care and diagnostic accuracy while minimizing radiation exposure.

ƏVƏZLİYİN TƏDRİSİNƏ İNTEQRATİV YANAŞMALAR

Integrative Approaches to Replacement Learning Summary A pronoun is a unique, peculiar part of speech. Pronouns prevent unnecessary repetition of words that denote objects, signs and qualities in speech. Pronouns replace and replace nouns, adjectives, numbers, and adverbs, allowing for deeper understanding of these parts of speech. Most definitions and rules learned during pronoun teaching have integrative features; Pronouns are also important when studying spelling and spelling material. When teaching pronouns in the elementary grades, careful attention should be paid to highlighting and integrating each topic with spelling and spelling, since learning the features of these topics is closely related to their spelling and pronuncia¬tion. When teaching pronouns, in addition to teaching their grammatical features, the rules of writing and pronunciation are also studied, which creates favorable conditions for the application of morphological knowledge and spelling and spelling rules together, students' literacy, increases their fluency and develops their speech. Incorporating pronouns into spelling and orthography results in rapid and robust learning of spelling and spelling rules. Key words: pronoun, learning, integration, spelling, spelling

The time course of stimulus-specific perceptual learning.

Practice on perceptual tasks can lead to long-lasting, stimulus-specific improvements. Rapid stimulus-specific learning, assessed 24hours after practice, has been found with just 105 practice trials in a face identification task. However, a much longer time course for stimulus-specific learning has been found in other tasks. Here, we examined 1) whether rapid stimulus-specific learning occurs for unfamiliar, non-face stimuli in a texture identification task; 2) the effects of varying practice across a range from just 21 trials up to 840 trials; and 3) if rapid, stimulus-specific learning persists over a 1-week, as well as a 1-day, interval. Observers performed a texture identification task in two sessions separated by one day (Experiment 1) or 1 week (Experiment 2). Observers received varying amounts of practice (21, 63, 105, or 840 training trials) in session 1 and completed 840 trials in session 2. In session 2, one-half of the observers in each group performed the task with the same textures as in session 1, and one-half switched to novel textures (same vs.novel conditions). In both experiments we found that stimulus-specific learning - defined as the difference in response accuracy in the same and novel conditions - increased as a linear function of the log number of session 1 training trials and was statistically significant after approximately 100 training trials. The effects of stimulus novelty did not differ across experiments. These results support the idea that stimulus-specific learning in our task arises gradually and continuously through practice, perhaps concurrently with general learning.

Probabilistic movement primitives based multi-task learning framework

With the increasing complexity of industrial production and manufacturing tasks, industrial robots are expected to learn intricate operations from simple actions easily and quickly with adaption to dynamic environment. In this paper, a task-parameterized multi-task learning framework is proposed to facilitate rapid learning of operational skills for industrial robots. In this framework, a conditional Probabilistic Movement Primitives (ProMP) is firstly employed to the single-task learning. Using the conditional probability calculation, the extrapolation issue in Learning from Demonstration (LfD) is addressed, enabling robots to learn beyond teaching. Subsequently, the single-task is extended to multi-task scenario by proposing a multi-task learning approach where each single task executes an extrapolation learning. The learned skill can meet the multiple task requirements through an iterative modulation manner. The effectiveness of the proposed framework is validated through both the simulation and a 7-DoF Franka-Emika robot experiment in a predefined task scenario. Furthermore, the outperformance of the proposed method is demonstrated by comparing with the state-of-art movement primitives based learning method.

Imperfect adaptation by freshwater crocodiles to the invasion of a toxic prey species

Novel interactions between invaders and native species that have evolved in their absence may impose strong selective pressures that drive species to extinction or prompt rapid co-evolution and learning. Here, we report on the effects that invasive cane toads, a toxic prey species, have had on freshwater crocodile populations in 7 waterholes of the Victoria River, Australia, before and up to 14 years after toads invaded. We recorded observations of crocodiles attacking toads, dissected dead crocodiles to determine if they had eaten toads and indexed the abundances of cane toads, live crocodiles and dead crocodiles. Following toad-invasion we observed crocodiles attacking cane toads. Dead crocodiles were only observed following the invasion of toads and 62% of the 71 dead crocodiles we dissected had toads in their stomachs. Counts of dead crocodiles showed a humped relationship with time since toad invasion and declined markedly after 3 years post-toad invasion. Live crocodile abundance declined sharply following toad-invasion, but this decline attenuated approximately 4 years post-invasion. The pulse of crocodile mortality and attenuation of the rate of crocodile population decline suggests that crocodiles have evolved or learned to enable co-existence with toads. However, crocodile populations have shown no sign of recovery in the 8–14 years post toad invasion. Our findings highlight that adaptation by native species to the presence of invaders may be imperfect and thus may not necessarily entail numerical recovery of populations to pre-invasion levels, but instead downward shifts to new equilibria due to ongoing interactions with invaders.

Machine learning and IoT based MIMO-UWB Antenna Integrated with Ku-Band for Seamless Wireless Communication

The implementation of machine learning algorithms with advanced antenna technologies, such as Multiple Input Multiple Output Ultra-Wideband -UWB- and Cup-Ban, has emerged as a powerful method for improving wireless communication systems. This research focuses on the effectiveness of applying three machine learning algorithms, which are Deep Q-Networks -DQN- , Label Propagation, and Model-Agnostic Meta-Learning (MAML). Using several experiments, we analyzed the throughput, latency, and packet loss of every algorithm. These were part of the project designed on a wireless signal combined with MIMO-UWB and Ku-Band. Nevertheless, MAML has the highest average throughput of 261.3 Mbps, and the other algorithms follow one by one: DQN with 252.5 Mbps and Label Propagation with 240.6 Mbps. As to latency, the lowest average is 9.0 ms and 11.0 ms for DQN and 14.2 ms for Label Propagation. Finally, MAML proved the least packet losses with the average value of 0.18% and DQN with 0.39% and 0.73% of Label Propagation .The results presented above indicate that by adapting more quickly to changing wireless environments, MAML effectively optimizes communication performance in terms of throughput, latency, and packet loss. Specifically, through its ability to accelerate the allocation of resources and management of networks as a whole, more frequent and rapid learning with MAML results in better system reliability and data integrity.

563 Team Science to Assess Effectiveness and Impact in Public Healthcare Delivery System Contracting

OBJECTIVES/GOALS: Healthcare organizations and payers are moving from accountability to effectiveness frameworks. Static vendor contracts for full-scale implementation limit organizations' ability to evaluate impact before scale-up, or to iteratively improve. Our team science innovation employs science and learning methods as systems engage vendors. METHODS/STUDY POPULATION: Our team science innovation is a method to assess and model impact of interventions at scale in healthcare delivery systems. We are integrating expertise in learning processes of an academic medical center (UCLA CTSI) with the organizational knowledge and methodological expertise of the nation’s largest Medicaid managed care plan (LA Care Health Plan), which has over 2 million members. The LA Care Advanced Analytics Lab has unique capability in machine learning, while enables deep learning of variation. Our innovative product is a template to quickly mobilize evaluation and learning for a diverse population in a varied and distributed delivery system. The template design enables rapid learning for the full-scale policy implementation often imposed by government, and in the short timeframes involved. RESULTS/ANTICIPATED RESULTS: LA Care and the UCLA CTSI partnered to provide subject matter expertise and design effective pilots for interventions such as transitional care services, complex care management, and physician home visit strategies, accounting for confounding factors affecting the intervention and outcome. So far, collaborative modeling and design has produced a successful pilot of a physician home visit program intended to reduce avoidable emergency department visits. This pilot quickly revealed several major changes that would need to be incorporated for the contracted vendor to produce results if operated at scale, further informed by machine learning, in sufficient time to inform the contracting process. There are multiple evolving applications, including housing/homelessness. DISCUSSION/SIGNIFICANCE: Integrating the large data and analytics of a large healthcare organization with learning methods from the CTSI -- including learning from variation and designs for studying impact during scale-up -- fosters academic-community team science that could significantly improve the value of our largest delivery systems, public and commercial.

The dark sugar factory – is it a feasible vision on the horizon?

The everyday discussion in industrial operations globally revolves around artificial intelligence, model-based control, and other similar digitally oriented topics. The question looms on whether a manpower-free sugar factory is possible one day. What is needed, and what is possible today? Even though evaporation crystallization in batch pans is a long-proven and fully automatic operation, demand is growing for a better understanding of how to operate a sugar house under optimized conditions. That is becoming more imperative as the “old hands” leave the factories and take experience with them. Experience is not easily replaced, but modern tools combining process automation with machine learning or AI components allow for a rapid learning curve and substantial support for day-to-day operations. This paper shines light on the need to not only investigate the parameters of the sugar house alone but that a plant-wide approach is needed to really understand what the right tuning points for optimized sugar house operations are

Examining The Acceptance of Tarannum Smart Learning Application Using Diffusion of Innovation Model

This paper presents a study aimed at developing a realistic model for understanding users' intention to use Smart Learning Application (SLA). SLA provides users with an intelligent learning platform that enables effective learning management, user-centered networking of intelligent devices, and rapid learning from any location. However, despite the advantages of SLA, its use is not widely accepted. The study used a quantitative approach, and data was collected from 229 respondents using a survey. SmartPLS (v. 3.2.8) was used to examine the relationships between the variables in the proposed model. The study developed a comprehensive model based on DOI that addressed users' intention to use SLA. The findings indicated that the direct influence of Relative Advantage, Compatibility, and Complexity on users' intention to use Tarannum Smart Learning Application (TSLA) was not significant, and hypotheses H1, H2, and H3 were not supported. However, hypothesis H4, which postulated a significant relationship between Observability and Intention to Use, was supported. The proposed model can be used as a basis for improving the learning of various types of tarannum by teachers teaching the holy Quran. The study can be expanded to other Smart Learning Applications to validate the findings and identify more opportunities. TSLA, by design, promotes human-oriented networking of smart devices and applications, making it a suitable tool for encouraging active users. Longitudinal data may be necessary to investigate the variables that impact people's acceptance level of TSLA. The technical implementation of TSLA from a policy maker's viewpoint can be explored in future studies.

Adaptive Meta-Learning Probabilistic Inference Framework for Long Sequence Prediction

Long sequence prediction has broad and significant application value in fields such as finance, wind power, and weather. However, the complex long-term dependencies of long sequence data and the potential domain shift problems limit the effectiveness of traditional models in practical scenarios. To this end, we propose an Adaptive Meta-Learning Probabilistic Inference Framework (AMPIF) based on sequence decomposition, which can effectively enhance the long sequence prediction ability of various basic models. Specifically, first, we decouple complex sequences into seasonal and trend components through a frequency domain decomposition module. Then, we design an adaptive meta-learning task construction strategy, which divides the seasonal and trend components into different tasks through a clustering-matching approach. Finally, we design a dual-stream amortized network (ST-DAN) to capture shared information between seasonal-trend tasks and use the support set to generate task-specific parameters for rapid generalization learning on the query set. We conducted extensive experiments on six datasets, including wind power and finance scenarios, and the results show that our method significantly outperforms baseline methods in prediction accuracy, interpretability, and algorithm stability and can effectively enhance the long sequence prediction capabilities of base models. The source code is publicly available at https://github.com/Zhu-JP/AMPIF.

Fast and Controllable Post-training Sparsity: Learning Optimal Sparsity Allocation with Global Constraint in Minutes

Neural network sparsity has attracted many research interests due to its similarity to biological schemes and high energy efficiency. However, existing methods depend on long-time training or fine-tuning, which prevents large-scale applications. Recently, some works focusing on post-training sparsity (PTS) have emerged. They get rid of the high training cost but usually suffer from distinct accuracy degradation due to neglect of the reasonable sparsity rate at each layer. Previous methods for finding sparsity rates mainly focus on the training-aware scenario, which usually fails to converge stably under the PTS setting with limited data and much less training cost. In this paper, we propose a fast and controllable post-training sparsity (FCPTS) framework. By incorporating a differentiable bridge function and a controllable optimization objective, our method allows for rapid and accurate sparsity allocation learning in minutes, with the added assurance of convergence to a predetermined global sparsity rate. Equipped with these techniques, we can surpass the state-of-the-art methods by a large margin, e.g., over 30\% improvement for ResNet-50 on ImageNet under the sparsity rate of 80\%. Our plug-and-play code and supplementary materials are open-sourced at https://github.com/ModelTC/FCPTS.

Focus-Then-Decide: Segmentation-Assisted Reinforcement Learning

Visual Reinforcement Learning (RL) is a promising approach to achieve human-like intelligence. However, it currently faces challenges in learning efficiently within noisy environments. In contrast, humans can quickly identify task-relevant objects in distraction-filled surroundings by applying previously acquired common knowledge. Recently, foundational models in natural language processing and computer vision have achieved remarkable successes, and the common knowledge within these models can significantly benefit downstream task training. Inspired by these achievements, we aim to incorporate common knowledge from foundational models into visual RL. We propose a novel Focus-Then-Decide (FTD) framework, allowing the agent to make decisions based solely on task-relevant objects. To achieve this, we introduce an attention mechanism to select task-relevant objects from the object set returned by a foundational segmentation model, and only use the task-relevant objects for the subsequent training of the decision module. Additionally, we specifically employed two generic self-supervised objectives to facilitate the rapid learning of this attention mechanism. Experimental results on challenging tasks based on DeepMind Control Suite and Franka Emika Robotics demonstrate that our method can quickly and accurately pinpoint objects of interest in noisy environments. Consequently, it achieves a significant performance improvement over current state-of-the-art algorithms. Project Page: https://www.lamda.nju.edu.cn/chenc/FTD.html Code: https://github.com/LAMDA-RL/FTD

Using a Rapid Learning Health System for Stratified Care in Emerging Adult Mental Health Services: Protocol for the Implementation of Patient-Reported Outcome Measures.

Mental illness among emerging adults is often difficult to ameliorate due to fluctuating symptoms and heterogeneity. Recently, innovative approaches have been developed to improve mental health care for emerging adults, including (1) implementing patient-reported outcome measures (PROMs) to assess illness severity and inform stratified care to assign emerging adults to a treatment modality commensurate with their level of impairment and (2) implementing a rapid learning health system in which data are continuously collected and analyzed to generate new insights, which are then translated to clinical practice, including collaboration among clients, health care providers, and researchers to co-design and coevaluate assessment and treatment strategies. The aim of the study is to determine the feasibility and acceptability of implementing a rapid learning health system to enable a measurement-based, stratified care treatment strategy for emerging adults. This study takes place at a specialty clinic serving emerging adults (age 16-24 years) in Calgary, Canada, and involves extensive collaboration among researchers, providers, and youth. The study design includes six phases: (1) developing a transdiagnostic platform for PROMs, (2) designing an initial stratified care model, (3) combining the implementation of PROMs with stratified care, (4) evaluating outcomes and disseminating results, (5) modification of stratified care based on data derived from PROMs, and (6) spread and scale to new sites. Qualitative and quantitative feedback will be collected from health care providers and youth throughout the implementation process. These data will be analyzed at regular intervals and used to modify the way future services are delivered. The RE-AIM (Reach, Effectiveness, Adoption, Implementation, and Maintenance) framework is used to organize and evaluate implementation according to 3 key objectives: improving treatment selection, reducing average wait time and treatment duration, and increasing the value of services. This project was funded through a program grant running from 2021 to 2026. Ethics approval for this study was received in February 2023. Presently, we have developed a system of PROMs and organized clinical services into strata of care. We will soon begin using PROMs to assign clients to a stratum of care and using feedback from youth and clinicians to understand how to improve experiences and outcomes. This study has key implications for researchers and clinicians looking to understand how to customize emerging adult mental health services to improve the quality of care and satisfaction with care. This study has significant implications for mental health care systems as part of a movement toward value-based health care. PRR1-10.2196/51667.

A Mobile Health App to Support Home-Based Aerobic Exercise in Neuromuscular Diseases: Usability Study.

Home-based aerobic exercise in people with neuromuscular diseases (NMDs) has benefits compared to exercise in the hospital or a rehabilitation center because traveling is often cumbersome due to mobility limitations, and societal costs are lower. Barriers to home-based aerobic exercise include reduced possibilities for monitoring and lack of motivation. To overcome these and other barriers, we developed a mobile health app: Keep on training with ReVi (hereafter referred to as ReVi). We aimed to determine the usability of the ReVi app. Patients followed a 4-month, polarized, home-based aerobic exercise program on a cycle or rowing ergometer, with 2 low-intensity sessions and 1 high-intensity session per week supported by the ReVi app. The app collected training data, including heart rate and ratings of perceived exertion, provided real-time feedback on reaching target intensity zones, and enabled monitoring via an online dashboard. Physiotherapists instructed patients on how to use the ReVi app and supervised them during their training program. Patients and physiotherapists separately evaluated usability with self-developed questionnaires, including 9 questions on a 5-point Likert scale, covering the usability elements efficiency, effectiveness, and satisfaction. Twenty-nine ambulatory adult patients (n=19 women; mean age 50.4, SD 14.2 years) with 11 different slowly progressive NMDs participated. Both patients and physiotherapists (n=10) reported that the app, in terms of its efficiency, was easy to use and had a rapid learning curve. Sixteen patients (55%) experienced 1 or more technical issue(s) during the course of the exercise program. In the context of effectiveness, 23 patients (81%) indicated that the app motivated them to complete the program and that it helped them to exercise within the target intensity zones. Most patients (n=19, 70%) and physiotherapists (n=6, 60%) were satisfied with the use of the app. The median attendance rate was 88% (IQR 63%-98%), with 76% (IQR 69%-82%) of time spent within the target intensity zones. Four adverse events were reported, 3 of which were resolved without discontinuation of the exercise program. The usability of the ReVi app was high, despite the technical issues that occurred. Further development of the app to resolve these issues is warranted before broader implementation into clinical practice.

Dynamic improvisation capabilities as a learning mechanism in early internationalizing firms

Our study addresses an inconsistency in the literature on whether a lack of knowledge in early internationalizing firms is an obstacle or an advantage. We integrate learning, capabilities, and improvisation literature to reveal how case firms from New Zealand and Finland internationalize early under uncertainty and time pressure. We develop a process model and propositions to show how firms develop improvisation capabilities and subsequently rapidly acquire the international business knowledge that can explain early internationalization. Improvisation and the associated rapid learning complement the learning advantage of newness with an alternative explanation for early internationalization.

Effects of training length on adaptation to noise-vocoded speech.

Listeners show rapid perceptual learning of acoustically degraded speech, though the amount of exposure required to maximize speech adaptation is unspecified. The current work used a single-session design to examine the length of auditory training on perceptual learning for normal hearing listeners exposed to eight-channel noise-vocoded speech. Participants completed short, medium, or long training using a two-alternative forced choice sentence identification task with feedback. To assess learning and generalization, a 40-trial pre-test and post-test transcription task was administered using trained and novel sentences. Training results showed all groups performed near ceiling with no reliable differences. For test data, we evaluated changes in transcription accuracy using separate linear mixed models for trained or novel sentences. In both models, we observed a significant improvement in transcription at post-test relative to pre-test. Critically, the three training groups did not differ in the magnitude of improvement following training. Subsequent Bayes factors analysis evaluating the test by group interaction provided strong evidence in support of the null hypothesis. For these stimuli and procedure, results suggest increased training does not necessarily maximize learning outcomes; both passive and trained experience likely supported adaptation. Findings may contribute to rehabilitation recommendations for listeners adapting to degraded speech signals.

Systematic Human Learning and Generalization From a Brief Tutorial With Explanatory Feedback.

We investigate human adults' ability to learn an abstract reasoning task quickly and to generalize outside of the range of training examples. Using a task based on a solution strategy in Sudoku, we provide Sudoku-naive participants with a brief instructional tutorial with explanatory feedback using a narrow range of training examples. We find that most participants who master the task do so within 10 practice trials and generalize well to puzzles outside of the training range. We also find that most of those who master the task can describe a valid solution strategy, and such participants perform better on transfer puzzles than those whose strategy descriptions are vague or incomplete. Interestingly, fewer than half of our human participants were successful in acquiring a valid solution strategy, and this ability was associated with completion of high school algebra and geometry. We consider the implications of these findings for understanding human systematic reasoning, as well as the challenges these findings pose for building computational models that capture all aspects of our findings, and we point toward a role for learning from instructions and explanations to support rapid learning and generalization.