Error Learning Research Articles

Learning and metacognition under volatility in GD: Lower learning rates and distorted coupling between action and confidence.

Decisions and learning processes are under metacognitive control, where confidence in one's actions guides future behaviour. Indeed, studies have shown that being more confident results in less action updating and learning, and vice versa. This coupling between action and confidence can be disrupted, as has been found in individuals with high compulsivity symptoms. Patients with Gambling Disorder (GD) have been shown to exhibit both higher confidence and deficits in learning. In this study, we tested the hypotheses that patients with GD display increased confidence, reduced action updating and lower learning rates. Additionally, we investigated whether the action-confidence coupling was distorted in patients with GD. To address this, 27 patients with GD and 30control participants performed a predictive inference task designed to assess action and confidence dynamics during learning under volatility. Action-updating, confidence and their coupling were assessed and computational modeling estimated parameters for learning rates, error sensitivity, and sensitivity to environmental changes. Contrary to our expectations, results revealed no significant group differences in action updating or confidence levels. Nevertheless, GD patients exhibited a weakened coupling between confidence and action, as well as lower learning rates. This suggests that patients with GD may underutilize confidence when steering future behavioral choices. Ultimately, these findings point to a disruption of metacognitive control in GD, without a general overconfidence bias in neutral, non-incentivized volatile learning contexts.

CREAD: A Classification-Restoration Framework with Error Adaptive Discretization for Watch Time Prediction in Video Recommender Systems

The watch time is a significant indicator of user satisfaction in video recommender systems. However, the prediction of watch time as a target variable is often hindered by its highly imbalanced distribution with a scarcity of observations for larger target values and over-populated samples for small values. State-of-the-art watch time prediction models discretize the continuous watch time into a set of buckets in order to consider the distribution of watch time. However, it is highly uninvestigated how these discrete buckets should be created from the continuous watch time distribution, and existing discretization approaches suffer from either a large learning error or a large restoration error. To address this challenge, we propose a Classification-Restoration framework with Error-Adaptive-Discretization (CREAD) to accurately predict the watch time. The proposed framework contains a discretization module, a classification module, and a restoration module. It predicts the watch time through multiple classification problems. The discretization process is a key contribution of the CREAD framework. We theoretically analyze the impacts of the discretization on the learning error and the restoration error, and then propose the error-adaptive discretization (EAD) technique to better balance the two errors, which achieves better performance over traditional discretization approaches. We conduct detailed offline evaluations on a public dataset and an industrial dataset, both showing performance gains through the proposed approach. Moreover, We have fully launched our framework to an online video platform, which resulted in a significant increase in users' video watch time by 0.29% through A/B testing. These results highlight the effectiveness of the CREAD framework in watch time prediction in video recommender systems.

Symmetric Q-learning: Reducing Skewness of Bellman Error in Online Reinforcement Learning

In deep reinforcement learning, estimating the value function to evaluate the quality of states and actions is essential. The value function is often trained using the least squares method, which implicitly assumes a Gaussian error distribution. However, a recent study suggested that the error distribution for training the value function is often skewed because of the properties of the Bellman operator, and violates the implicit assumption of normal error distribution in the least squares method. To address this, we proposed a method called Symmetric Q-learning, in which the synthetic noise generated from a zero-mean distribution is added to the target values to generate a Gaussian error distribution. We evaluated the proposed method on continuous control benchmark tasks in MuJoCo. It improved the sample efficiency of a state-of-the-art reinforcement learning method by reducing the skewness of the error distribution.

The role of momentary emotions in promoting error learning orientation among lower secondary school students: An intervention study embedded in a short visual programming course.

Students make many errors in visual programming. In order to learn from these, it is important that students regulate their emotions and view errors as learning opportunities. This study aimed to explore to what extent momentary emotions, specifically enjoyment, anxiety and boredom, as well as the error learning orientation of students, interacted during a 1-day course on visual programming in an out-of-school learning environment. The sample consisted of 269 lower secondary school students (grades 7-9). The data were collected in an intervention study, with questionnaires applied directly before and after the course, and with four measurements of state emotions during the course. The results showed that error learning orientation had an expected effect on the students' emotions at the beginning of the course. The emotions changed positively over the course of the workshop, while the error learning orientation remained stable. No differences in error learning orientation were found between the control and intervention groups. An expected, reciprocal effect of students' emotions on their error learning orientation at the end of the course day could not be found. Changes in error learning orientation are difficult to achieve during 1-day courses. Nevertheless, through targeted, pedagogical approaches, which aim to minimize the influence of unfavourable emotions that occur in problem-oriented learning situations, teachers could help students develop a positive error learning orientation in the long term, whereby errors are viewed as an opportunity for learning.

Examining Patterns and Predictors of ADHD Teens' Skill-Learning Trajectories During Enhanced FOrward Concentration and Attention Learning (FOCAL+) Training.

Examine patterns and predictors of skill learning during multisession Enhanced FOrward Concentration and Attention Learning (FOCAL+) training. FOCAL+ teaches teens to reduce the duration of off-road glances using real-time error learning. In a randomized controlled trial, teens with ADHD received five sessions of FOCAL+ training and demonstrated significant reductions in extended glances (>2-s) away from the roadway (i.e., long-glances) and a 40% reduced risk of a crash/near-crash event. Teens' improvement in limiting long-glances as assessed after each FOCAL+ training session has not been examined. Licensed teen (ages 16-19) drivers with ADHD (n = 152) were randomly assigned to five sessions of either FOCAL+ or modified standard driver training. Teens completed driving simulation assessments at baseline, after each training session, and 1 month and 6 months posttraining. Naturalistic driving was monitored for one year. FOCAL+ training produced a 53% maximal reduction in long-glances during postsession simulated driving. The number of sessions needed to achieve maximum performance varied across participants. However, after five FOCAL+ training sessions, number of long-glances was comparable irrespective of when teens achieved their maximum performance. The magnitude of reduction in long-glances predicted levels of long-glances during simulated driving at 1 month and 6 months posttraining but not naturalistic driving outcomes. FOCAL+ training provided the most benefit during training to teens who were younger and had less driving experience. FOCAL+ training significantly reduces long-glances beginning at the 1st training session. Providing five FOCAL+ training sessions early on during teen driving may maximize benefit.

Generative Quantum Machine Learning via Denoising Diffusion Probabilistic Models.

Deep generative models are key-enabling technology to computer vision, text generation, and large language models. Denoising diffusion probabilistic models (DDPMs) have recently gained much attention due to their ability to generate diverse and high-quality samples in many computer vision tasks, as well as to incorporate flexible model architectures and a relatively simple training scheme. Quantum generative models, empowered by entanglement and superposition, have brought new insight to learning classical and quantum data. Inspired by the classical counterpart, we propose the quantum denoising diffusion probabilistic model (QuDDPM) to enable efficiently trainable generative learning of quantum data. QuDDPM adopts sufficient layers of circuits to guarantee expressivity, while it introduces multiple intermediate training tasks as interpolation between the target distribution and noise to avoid barren plateau and guarantee efficient training. We provide bounds on the learning error and demonstrate QuDDPM's capability in learning correlated quantum noise model, quantum many-body phases, and topological structure of quantum data. The results provide a paradigm for versatile and efficient quantum generative learning.

Reinforcement Learning-Based Fractional-Order Adaptive Fault-Tolerant Formation Control of Networked Fixed-Wing UAVs With Prescribed Performance.

This article investigates the fault-tolerant formation control (FTFC) problem for networked fixed-wing unmanned aerial vehicles (UAVs) against faults. To constrain the distributed tracking errors of follower UAVs with respect to neighboring UAVs in the presence of faults, finite-time prescribed performance functions (PPFs) are developed to transform the distributed tracking errors into a new set of errors by incorporating user-specified transient and steady-state requirements. Then, the critic neural networks (NNs) are developed to learn the long-term performance indices, which are used to evaluate the distributed tracking performance. Based on the generated critic NNs, actor NNs are designed to learn the unknown nonlinear terms. Moreover, to compensate for the reinforcement learning errors of actor-critic NNs, nonlinear disturbance observers (DOs) with skillfully constructed auxiliary learning errors are developed to facilitate the FTFC design. Furthermore, by using the Lyapunov stability analysis, it is shown that all follower UAVs can track the leader UAV with predesigned offsets, and the distributed tracking errors are finite-time convergent. Finally, comparative simulation results are presented to show the effectiveness of the proposed control scheme.

Discovery and learning of feature geometry in underwater acoustic sensing: Case studies and challenges in active sonar applications

Interpretable artificial intelligence (AI) and related machine learning (ML) techniques are gaining popularity for underwater acoustic sensing applications. However, interpretability of machine learnt features poses a fundamental challenge to successful application of these powerful data science techniques to underwater acoustics. Autonomous sonar target recognition is especially interesting from the data science perspective as beyond target-specific features, a sonar ping response typically includes significant interference from the environment, e.g., clutter, multipath scattering, etc. Such environmental interference, resulting from complex, dynamic, unpredictable, and often unknown factors, manifest as identifiable structures in acoustic color or alternate multi-dimensional feature representations that can lead to machine learning and classification errors. In this talk, we will discuss these challenges commonly encountered in underwater acoustics using case studies from field experiments in active sonar target recognition. Some results will also include physics-driven simulations in this domain to provide robust ground truths. In particular, we will posit how braided feature geometry and its representation, as well geometry of overlap between features can render a sonar target feature informative, explainable, and discoverable. [Work funded by the ONR Grant Nos. N000142112420 and N000142312503 and DoD Navy (NEEC) Grant No. N001742010016.]

Dissimilarities of neural representations of extinction trials are associated with extinction learning performance and renewal level.

Renewal of extinguished responses is associated with higher activity in specific extinction-relevant brain regions, i.e., hippocampus (HC), inferior frontal gyrus (IFG), and ventromedial PFC (vmPFC). HC is involved in processing of context information, while IFG and vmPFC use such context information for selecting and deciding among competing response options. However, it is as yet unknown to what extent trials with changed versus unchanged outcome, or extinction trials that evoke renewal (i.e., extinction context differs from acquisition and test context: ABA trials) and trials that do not (i.e., same context in all phases: AAA trials) are represented differentially in extinction-relevant brain regions. In this study, we applied representational similarity analysis (RSA) to determine differences in neural representations of these trial types and their relationship to extinction error rates and renewal level. Overall, individuals with renewal (REN) and those without (NoREN) did not differ significantly in their discrimination levels between ABA and AAA extinction trials, with the exception of right posterior HC, where REN exhibited more pronounced context-related discrimination. In addition, higher dissimilarity of representations in bilateral posterior HC, as well as in several IFG regions, during extinction learning was linked to lower ABA renewal rates. Both REN and NoREN benefitted from prediction error feedback from ABA extinction errors for context- and outcome-related discrimination of trials in IFG, vmPFC, and HC, but only the NoREN group also benefitted from error feedback from AAA extinction errors. Thus, while in both groups the presence of a novel context supported formation of distinct representations, only in NoREN the expectancy violation of the surprising change of outcome alone had a similar effect. In addition, only in NoREN context-related discrimination was linked to error feedback in vmPFC. In summary, the findings show that context- and outcome-related discrimination of trials in HC, vmPFC, and IFG is linked to extinction learning errors, regardless of renewal propensity, and at the same time point towards differential context processing strategies in REN and NoREN. Moreover, better discrimination of context-related trials during extinction learning promotes less renewal during extinction recall, suggesting that renewal may be related to suboptimal context-related trial discrimination.

Organizational learning, quality and safety culture in patient care: comparing Colombia and the Kingdom of Saudi Arabia

PurposeThe purpose of this study is to determine the interactions between factors such as organizational learning, feedback about errors, punitive response to errors and communication quality in hospitals in the Kingdom of Saudi Arabia and Colombia when adopting a culture of quality and safety (CQS) in patient care.Design/methodology/approachBased on a literature review, a self-administered questionnaire was developed and used to collect data from 417 Saudi respondents affiliated with hospitals and 483 Colombian respondents at the beginning of the pandemic. Structural equation modeling is used in this study to test the hypothesized relationships.FindingsThe results show a solid and significant predictive relationship between feedback about errors and the CQS in both countries (Colombia: b = 0.55, p < 0.001; KSA: b = 0.44, p < 0.001), but a very low and insignificant predictive relationship between no punitive response to errors and CQS (Colombia: b = –0.02, p > 0.05; KSA: b = 0.05, p > 0.05).Practical implicationsThis study demonstrates the importance of organizational learning in fostering a CQS in the health-care sector in the Kingdom of Saudi Arabia and Colombia. Recent unprecedented policy actions motivated by the COVID-19 pandemic, such as social distancing, lockdowns and safety practices enforcement, have further highlighted this concern. Moreover, attention to the dimensions addressed in this study is required for accreditation purposes in organizations seeking to promote a CQS. Overall, this research highlights the vital role of safety and quality practices among health-care organizations, which has significant policy implications, especially in the current period of high uncertainty.Originality/valueThis paper contributes to the theory and practice in the health-care sector by extending the current knowledge of the impact of the quality of communications, non-punitive response to errors and feedback about errors in organizational learning and safety culture, and by presenting a novel, quantitative methodology seldom used for these topics.

Preserving the self with artificial intelligence using VIPCare-a virtual interaction program for dementia caregivers.

Assistive technology is increasingly used to support the physical needs of differently abled persons but has yet to make inroads on support for cognitive or psychological issues. This gap is an opportunity to address another-the lack of contribution from theoretical social science that can provide insights into problems that cannot be seen. Using Affect Control Theory (ACT), the current project seeks to close that gap with an artificially intelligent application to improve interaction and affect for people with Alzheimer's Disease and Related Dementias (ADRD). Using sociological theory, it models interactions with persons with ADRD based on self-sentiments, rather than cognitive memory, and informs a cellphone-based assistive tool called VIPCare for supporting caregivers. Staff focus groups and interviews with family members of persons with ADRD in a long-term residential care facility collected residents' daily needs and personal histories. Using ACT's evaluation, potency, and activity dimensions, researchers used these data to formulate a self-sentiment profile for each resident and programmed that profile into the VIPCare application. VIPCare used that profile to simulate affectively intelligent social interactions with each unique resident that reduce deflection from established sentiments and, thus, negative emotions. We report on the data collection to design the application, develop self-sentiment profiles for the resident, and generate assistive technology that applies a sociological theory of affect to real world management of interaction, emotion, and mental health. By reducing trial and error in learning to engage people with dementia, this tool has potential to smooth interaction and improve wellbeing for a population vulnerable to distress.

Byzantine-robust decentralized stochastic optimization with stochastic gradient noise-independent learning error

This paper studies Byzantine-robust stochastic optimization over a decentralized network, where every agent periodically communicates with its neighbors to exchange local models, and then updates its own local model with one or a mini-batch of local samples. The performance of such a method is affected by an unknown number of Byzantine agents, which conduct adversarially during the optimization process. To the best of our knowledge, there is no existing work that simultaneously achieves a linear convergence speed and a small learning error. We observe that the unsatisfactory trade-off between convergence speed and learning error is due to the intrinsic stochastic gradient noise. Motivated by this observation, we introduce two variance reduction methods, stochastic average gradient algorithm (SAGA) and loopless stochastic variance-reduced gradient (LSVRG), to Byzantine-robust decentralized stochastic optimization for eliminating the negative effect of the stochastic gradient noise. The two resulting methods, BRAVO-SAGA and BRAVO-LSVRG, enjoy both linear convergence speeds and stochastic gradient noise-independent learning errors. Such learning errors are optimal for a class of methods based on total variation (TV)-norm regularization and stochastic subgradient update. We conduct extensive numerical experiments to show their effectiveness under various Byzantine attacks.

Diet patterns and cognitive performance in a UK Female Twin Registry (TwinsUK)

BackgroundPlant-based diets may provide protection against cognitive decline and Alzheimer’s disease, but observational data have not been consistent. Previous studies include early life confounding from socioeconomic conditions and genetics that are known to influence both cognitive performance and diet behaviour. This study investigated associations between Mediterranean (MED) diet and MIND diets and cognitive performance accounting for shared genotype and early-life environmental exposures in female twins.MethodsDiet scores were examined in 509 female twins enrolled in TwinsUK study. The Cambridge Neuropsychological Test Automated Battery was used to assess cognition at baseline and 10 years later (in n = 275). A co-twin case–control study for discordant monozygotic (MZ) twins examined effects of diet on cognitive performance independent of genetic factors. Differences in relative abundance of taxa at 10-year follow-up were explored in subsamples.ResultsEach 1-point increase in MIND or MED diet score was associated with 1.75 (95% CI: − 2.96, − 0.54, p = 0.005 and q = 0.11) and 1.67 (95% CI: − 2.71, − 0.65, p = 0.002 and q = 0.02) fewer respective errors in paired-associates learning. Within each MZ pair, the twin with the high diet score had better preservation in spatial span especially for MED diet (p = 0.02). There were no differences between diet scores and 10-year change in the other cognitive tests. MIND diet adherence was associated with higher relative abundance of Ruminococcaceae UCG-010 (0.30% (95% CI 0.17, 0.62), q = 0.05) which was also associated with less decline in global cognition over 10 years (0.22 (95% CI 0.06, 0.39), p = 0.01).ConclusionsMIND or MED diets could help to preserve some cognitive abilities in midlife, particularly episodic and visuospatial working memory. Effects may be mediated by high dietary fibre content and increased abundance of short-chain fatty acid producing gut bacteria. Longer follow-up with repeated measures of cognition will determine whether diet can influence changes in cognition occurring in older age.

Unraveling Digital Transformation Dynamics in Manufacturing: A Mediation and Moderation Analysis

This study investigates the mediating and moderating mechanisms shaping the relationship between digital orientation and digital transformation performance in manufacturing enterprises. Drawing on data from 393 firms, it confirms the positive influence of digital orientation on acquiring learning and trial and error learning, which in turn mediate the relationship between digital orientation and digital transformation performance. Furthermore, the study reveals the moderating roles of digital infrastructure and organizational agility in these relationships. The results highlight the pivotal role of organizational learning processes and technological infrastructure in fostering successful digital transformations within manufacturing contexts. By shedding light on these intricate dynamics, this research contributes to a deeper understanding of how organizations can effectively leverage digital resources to drive innovation and competitiveness in today's rapidly evolving business landscape.

Feed Concentration Forecasting Using Closed-Loop Input Error and Deep Learning

Feed Concentration Forecasting Using Closed-Loop Input Error and Deep Learning

Special major 1, 3 satisfiability logic in discrete Hopfield neural networks

&lt;abstract&gt; &lt;p&gt;Currently, the discrete Hopfield neural network deals with challenges related to searching space and limited memory capacity. To address this issue, we propose integrating logical rules into the neural network to regulate neuron connections. This approach requires adopting a specific logic framework that ensures the network consistently reaches the lowest global energy state. In this context, a novel logic called major 1,3 satisfiability was introduced. This logic places a higher emphasis on third-order clauses compared to first-order clauses. The proposed logic is trained by the exhaustive search algorithm, aiming to minimize the cost function toward zero. To evaluate the proposed model effectiveness, we compare the model's learning and retrieval errors with those of the existing non-systematic logical structure, which primarily relies on first-order clauses. The similarity index measures the similarity benchmark neuron state with the existing and proposed model through extensive simulation studies. Certainly, the major random 1,3 satisfiability model exhibited a more extensive solution space when the ratio of third-order clauses exceeds 0.7% compared to first-order clauses. As we compared the experimental results with other state-of-the-art models, it became evident that the proposed model achieved significant results in capturing the overall neuron state. These findings emphasize the notable enhancements in the performance and capabilities of the discrete Hopfield neural network.&lt;/p&gt; &lt;/abstract&gt;

An Error Learning Scenario-Based Scheme to Quantized Identification in Wiener-Hammerstein Systems Subject to Deadzone Nonlinearity

An Error Learning Scenario-Based Scheme to Quantized Identification in Wiener-Hammerstein Systems Subject to Deadzone Nonlinearity

A Nonlinear PID-Incorporated Adaptive Stochastic Gradient Descent Algorithm for Latent Factor Analysis

High-dimensional and incomplete (HDI) data are commonly encountered in various big data-related applications concerning the complex interactions among numerous nodes, such as the user-item iterations in a recommender system. A stochastic gradient descent (SGD)-based latent factor analysis (LFA) model can perform efficient representation learning to such HDI data, thereby extracting useful knowledge from them. However, a standard SGD algorithm updates a latent factor based on the current stochastic gradient only, without the considerations on the past information, making a resultant model suffer from slow convergence. To address this critical issue, this paper proposes an Adaptive Non-linear PID-incorporated SGD (ANPS) algorithm with two-fold ideas: 1) rebuilding the instant learning error when computing the stochastic gradient following the principle of a nonlinear PID controller to incorporate past update information into the learning scheme efficiently, and 2) implementing gain parameter adaptation following the principle of particle swarm optimization (PSO). Experiments on six widely-adopted HDI datasets demonstrate that compared with state-of-the-art LFA models, an ANPS-based LFA model achieves significant advantage in both efficiency and accuracy. Moreover, its flexible gain parameter adaptation mechanism greatly boosts its practicability for real issues. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —In many industrial applications like recommender systems, social network systems, and cloud service systems, people usually encounter numerous nodes and their highly-incomplete relationships. An HDI matrix is commonly adopted to describe such specific relationships. One of the major challenges is to acquire useful knowledge from an HDI matrix efficiently and accurately for various data analysis tasks, e.g., accurate recommendation, community detection, and web service selection. An SGD-based LFA model has been widely adopted to tackle this issue. However, it suffers from slow convergence that leads to considerable time cost on large-scale datasets. This study proposes an ANPS algorithm following the principle of a nonlinear PID controller. With it, an ANPS-based LFA model is achieved, which possesses fast convergence rate on an industrial HDI matrix. The proposed ANPS algorithm can be leveraged for different types of various machine learning models, thereby improving their utility and scalability in practice.

Research on the application of neural networks in macroeconomic forecasting and strategic management decision-making

Abstract Traditional macroeconomic forecasting models have some limitations in utilizing large-scale variables, screening appropriate variables, and forecasting accuracy. Based on the theories of economic fluctuation and New Keynes, this paper uses monetary policy, investment behavior, and consumption behavior as predictors of macroeconomic fluctuation. Then, it explores the principles and steps of the BP network applied to forecasting, the design of BP neural network structure, the selection of training algorithm and training parameters, and finally establishes a reasonable BP network structure model. Finally, using the prediction of GDP growth rate as an example, an empirical comparative analysis of model prediction accuracy is carried out. The empirical results of the model in the training set and the test set show that the structure of the BP neural network model is 3-48-1, and the model obtains the optimal learning error (6.207×10−5) at 500 iterations. In the experimental set, the root-mean-square prediction error between the output predicted value and the actual value is less than ±0.2 . The model accurately predicts the GDP growth rate, which can provide a theoretical basis for proposing macroeconomic strategic management decisions.

Analysis of the application of neural network technologies in predicting equipment failure in production

The relevance. The article considers the solution to the problem of analyzing and predicting changes in the dynamics of measuring instrument readings and identifying patterns describing the course and possible periods of failures of technological process equipment by constructing a predictive time series model. A critically important infrastructure facility of a thermal power plant was selected as the studied one – a water pump for pumping process water into neutralizer tanks (the first stage of chemical water purification) located in the chemical workshop of the Krasnoyarsk CHP-3. The general methodology of model construction, requirements for input data arrays, preprocessing algorithms for the formation of samples used for training and testing models are described. To build a predictive model, the work used elements such as recurrent layers of long-term short- term memory LSTM, MinMaxScaler regularization, Dropout layer to reduce the effect of retraining, the ‘Adam’ optimizer using error back propagation methods and optimization using the steepest descent method using conjugate gradients, the learning error function is the average root of the error (Root Mean Squared Error, RMSE), as well as plot_predictions visualization. The algorithms are implemented in Python, in the Anaconda Jupyter Nootebook core. Based on the results of testing the training, the model showed sufficient performance of this method, since the time spent on creating the model and obtaining a forecast plays an important role when launching the model in real production conditions. In conclusion, recommendations are given to improve the accuracy of the forecast, the developed interface of the production system is presented and the directions of future research in this area are outlined.