Improved Learning Rate Research Articles

BPNN model based AI for the estimation of soot data from flame luminosity emissions in H2/N2 diluted ethylene laminar diffusion flames

Advanced combustion devices and alternative fuels like hydrogen require an estimation of the impact of such changes on engine exhaust emissions of regulated pollutants like soot. Huge experimental data are then required to be collected and post-processed, both in academic and industrial setups. Machine learning and artificial intelligence can be employed to reduce cost and time. This study presents a Back Propagation Neural Network (BPNN) model that has been validated and optimized for the simultaneous prediction of soot volume fraction, temperature and particle sizes from flame luminosity measurements. The compatibility, robustness and reliability of neural network models in terms of fuel composition and flow rate variations was analyzed , by varying both fuel flow rate and fuel stream dilution in axisymmetric diffusion flame burning in a coflow of air. BPNN model predicted results were contrasted against those obtained through laser diagnostics. A detailed analysis of the performance of BPNN model, based on an extensive experimental study revealed indifference to N2/H2 dilution and fuel flow rate variations. Furthermore, an original BPNN model is optimized to improve learning rate and to reduce computation cost, by applying a Bayesian algorithm which continuously monitors and minimizes an error function. It was concluded that a training set size of three was sufficient to predict soot field parameters with a fair prediction accuracy. These results have direct implication for the application of proposed technique in practical combustion equipment where operating parameter, e.g. Air–fuel ratio and EGR, show both gradual and instantaneous variations and affect both performance and emissions. These results clearly demonstrate the robustness of this technique which offers opportunities for real-time, in situ, monitoring of practical combustion devices. Hence, Real-time data of sooting characteristics of flames can be used to both monitor and control pollutant emissions by integrating a response strategy in the system.

A comprehensive approach for PV wind forecasting by using a hyperparameter tuned GCVCNN-MRNN deep learning model

Precise power forecasting is important in contemporary energy management systems, especially for optimizing the use of renewable resources. However, precise predictions of solar and wind power face challenges due to their dependence on various environmental factors. In this research, we introduce a novel hybrid model that merges Convolutional Neural Network (CNN) and ResNet architectures with a Modified Recurrent Neural Network (MRNN). This MRNN model incorporates several key layers, including Bidirectional Long Short-Term Memory (BiLSTM), LSTM, and Bidirectional weighted Gated Recurrent Unit (BiGRU) layers. to address the constraints of conventional estimation methods and achieve precise short-term Photovoltaic and wind power prediction. Hyperparameters of the proposed CNN-MRNN model are optimized using the Grid Search CV algorithm, resulting in improved learning rate and performance. Comparative analysis with GCVCNN-RNN, GCVCNN-BiGRU,CNN-BiLSTM, and CNN-LSTM models demonstrates the superior performance of the proposed technique, achieving a 20% low mean absolute error (MAE) and 30% of root mean square error (RMSE). Additionally, the GCVCNN-MRNN model outperforms existing techniques based on Nash Sutcliffe analysis (0.9852) for time series forecasting. The experimental results validate the usefulness of the proposed GCVCNN-MRNN in correctly predicting PV and wind power, the model consistently achieves lower MAE and RMSE, along with higher correlation coefficient (CC) and R2, indicating improved accuracy and stability. This research contributes to the advancement of clean energy and expands the existing body of knowledge in power forecasting.

Learning Rates for Nonconvex Pairwise Learning

Pairwise learning is receiving increasing attention since it covers many important machine learning tasks, e.g., metric learning, AUC maximization, and ranking. Investigating the generalization behavior of pairwise learning is thus of great significance. However, existing generalization analysis mainly focuses on the convex objective functions, leaving the nonconvex pairwise learning far less explored. Moreover, the current learning rates of pairwise learning are mostly of slower order. Motivated by these problems, we study the generalization performance of nonconvex pairwise learning and provide improved learning rates. Specifically, we develop different uniform convergence of gradients for pairwise learning under different assumptions, based on which we characterize empirical risk minimizer, gradient descent, and stochastic gradient descent. We first establish learning rates for these algorithms in a general nonconvex setting, where the analysis sheds insights on the trade-off between optimization and generalization and the role of early-stopping. We then derive faster learning rates of order O(1/n) for nonconvex pairwise learning with a gradient dominance curvature condition, where n is the sample size. Provided that the optimal population risk is small, we further improve the learning rates to O(1/n2), which, to the best of our knowledge, are the first O(1/n2) rates for pairwise learning.

Interpretability-Guided Inductive Bias For Deep Learning Based Medical Image.

Deep learning methods provide state of the art performance for supervised learning based medical image analysis. However it is essential that trained models extract clinically relevant features for downstream tasks as, otherwise, shortcut learning and generalization issues can occur. Furthermore in the medical field, trustability and transparency of current deep learning systems is a much desired property. In this paper we propose an interpretability-guided inductive bias approach enforcing that learned features yield more distinctive and spatially consistent saliency maps for different class labels of trained models, leading to improved model performance. We achieve our objectives by incorporating a class-distinctiveness loss and a spatial-consistency regularization loss term. Experimental results for medical image classification and segmentation tasks show our proposed approach outperforms conventional methods, while yielding saliency maps in higher agreement with clinical experts. Additionally, we show how information from unlabeled images can be used to further boost performance. In summary, the proposed approach is modular, applicable to existing network architectures used for medical imaging applications, and yields improved learning rates, model robustness, and model interpretability.

Flying by Feeling: Communicating Flight Envelope Protection through Haptic Feedback

ABSTRACT Modern aircraft can be equipped with a flight envelope protection system: automation which modifies pilot control inputs to ensure that the aircraft remains within the allowable limits. Overruling the pilot inputs may lead to mode confusion, even when visual or auditory feedback is provided to alert pilots. We advocate using active control devices to make the flight envelope protection system tangible to the pilot. This paper presents the main findings of an evaluation of three haptic feedback designs for flight envelope protection. The first concept used both force feedback and vibro-tactile alerts, producing promising, yet inconclusive, results. The second concept used asymmetric vibrations to give directional alerting cues, which did not result in improved performance on initial use, but which did yield improved learning rate for the task. The third system employed force feedback to physically guide the pilot away from flight envelope limits, which yielded safety improvements from the first use, but created dependence: pilot performance degraded immediately after the force feedback was removed. From this, we advise to use asymmetric vibrations during training for flight envelope excursions, to leverage active control interfaces for providing force feedback during operation, and reevaluate a combination of both to combine their advantages for single-pilot operations.

On the improvement of accommodation to non-individual HRTFs via VR active learning and inclusion of a 3D room response

This study examines the efficiency of a training protocol using a virtual reality application designed to accelerate individual’s selection of, and accommodation to, non-individualized HRTF profiles. This training introduces three elements to hasten audio localization performance improvement: an interactive HRTF selection method, a parametric training program based on active learning, and a relatively dry room acoustic simulation designed to increase the quantity of spatial cues presented. Participants rapidly selected an HRTF (≈5 min) followed by training over three sessions of 12 min distributed over 5 days. To study the impact of the room acoustic component on localization performance evolution, participants were divided into two groups: one acting as control reference, training with only anechoic renderings, the other training in reverberant conditions. The efficiency of the training program was assessed across groups and the entire protocol was assessed through direct comparisons with results reported in previous studies. Results indicate that the proposed training program led to improved learning rates compared to that of previous studies, and that the included room response accelerated the learning process.

Construction and Simulation of Financial Audit Model Based on Convolutional Neural Network.

Big data has brought a new round of information revolution. Faced with the goal of full coverage of audit and supervision, making full use of big data is the main method to promote the realization of the goal of full coverage of audit and supervision. Data analysis and utilization is an indispensable task of auditing. Actively exploring multidimensional and intelligent data analysis methods and developing big data audit cases are the new development direction of auditing. The convolutional neural network's excellent ability to extract data features well meets the relevant requirements of financial auditing. However, in practical applications, convolutional neural networks often encounter various problems such as disappearance of gradients and difficulty in convergence, which reduces its expected performance in financial audit applications. In order to make the performance of the financial audit model based on convolutional neural network more excellent, after summarizing the characteristics of genetic algorithm, this article applies genetic algorithm to the optimization of the convolutional neural network model. We applied genetic algorithm to optimize the initial weights of the convolutional neural network. The error sensitivity and learning rate changes of different hidden layers are discussed, the influence of different learning rates on the convergence speed of convolutional neural networks is analyzed, and the recognition performance of other algorithms on financial audit data sets is simulated and compared. We conducted experiments on the network structure and parameter optimization on the financial audit database. The results show that the recognition error rate of the convolutional neural network model with improved learning rate algorithm in the financial audit data set is lower than that of the multilayer feedforward network, so it has better performance.

The Effectiveness of Public Investment in Basic Education in Brazil

The analysis of educational systems involves four central dimensions: economic, pedagogical, political and cultural. This paper focuses on the economic perspective, which has gained strength with the creation of the Basic Education Development Index (Ideb), and facilitates comparative analyzes of the performance of Brazilian educational systems. This research uses a linear least squares regression in which the dependent variables were the Ideb scores of Brazilian states and the independent variables were the corresponding investments in the maintenance and development of education (MDE), in the period 2005-2015. The results allow a comparative picture to be drawn of the effectiveness of the use of state resources invested by the states. It is verified that the investments in basic education in the states only partially explain the improved learning rates measured by the Ideb. Three distinct situations were observed. In 23 states there was an increase in investment in education accompanied by improvement in the Ideb. In 18 states, the increase in investments accompanied the improvement in the Ideb over a certain period, although this effect was not sustained throughout the historical series. In only 05 states, the relationship between investment and improvement in the Ideb remained positive throughout the analyzed period.

Neurally underdeveloped cuttlefish newborns exhibit social learning.

Learning can occur through self-experience with the environment, or through the observation of others. The latter allows for adaptive behaviour without trial-and-error, thus maximizing individual fitness. Perhaps given their mostly solitary lifestyle, cuttlefish have seldomly been tested under observational learning scenarios. Here we used a multi-treatment design to disentangle if and how neurally immature cuttlefish Sepia officinalis hatchlings (up to 5days) incorporate social information into their decision-making, when performing a task where inhibition of predatory behaviour is learned. In the classical social learning treatment using pre-trained demonstrators, observers did not register any predatory behaviour. In the inhibition by social learning treatment, using naïve (or sham) demonstrators, more observers than demonstrators learned the task, while also reaching learning criterion in fewer trials, and performing less number of attacks per trial. Moreover, the performance of demonstrator-observer pairs was highly correlated, indicating that the mere presence of conspecifics did not explain our results by itself. Additionally, observers always reported higher latency time to attack during trials, a trend that was reversed in the positive controls. Lastly, pre-exposure to the stimulus did not improve learning rates. Our findings reveal the vicarious capacity of these invertebrate newborns to learn modulation (inhibition) of predatory behaviour, potentially through emulation (i.e. affordance learning). Despite ongoing changes on neural organization during early ontogeny, cognitively demanding forms of learning are already present in cuttlefish newborns, facilitating behavioural adaptation at a critical life stage, and potentially improving individual fitness in the environment.

A New Approach to Distributed Hypothesis Testing and Non-Bayesian Learning: Improved Learning Rate and Byzantine Resilience

We study a setting where a group of agents, each receiving partially informative private signals, seek to collaboratively learn the true underlying state of the world (from a finite set of hypotheses), which generates their joint observation profiles. To solve this problem, we propose a distributed learning rule that differs fundamentally from existing approaches, in that it does not employ any form of “belief-averaging.” Instead, agents update their beliefs based on a min-rule. Under standard assumptions on the observation model and the network structure, we establish that each agent learns the truth asymptotically almost surely. As our main contribution, we prove that with probability 1, each false hypothesis is ruled out by every agent exponentially fast, at a network-independent rate that is strictly larger than existing rates. We then develop a computationally efficient variant of our learning rule that is provably resilient to agents who do not behave as expected (as represented by a Byzantine adversary model) and deliberately try to spread misinformation.

Neurophysiological and behavioural effects of tDCS upon memory and learning in schizophrenia

Schizophrenia is associated with significant impairment of cognitive function in the domains of memory and learning, including working memory (WM), implicit (IL) and probabilistic learning (PL). Previous research suggests that tDCS stimulation can improve cognitive functioning. We investigated if cognitive training with adjunct tDCS would improve learning rates in individuals with schizophrenia. The mechanism of action was investigated during an fMRI visit with concurrent tDCS. 49 right-handed patients with schizophrenia were randomized into real or sham tDCS. Participants completed 8 sessions of cognitive training, twice a day spread over 4 days (1, 2, 14 and 56). Anodal tDCS to the left dorso-lateral prefrontal cortex (DLPC) was applied during day 1 and day 14. The cognitive training included a WM, IL, and PB task. Those suitable and willing underwent an fMRI scan on day 14, during which they completed an additional executive function (EF) task. Multilevel regressions were used to compare the learning rates between the real and sham tDCS groups. General linear models were used to analyse the fMRI data. We found a delayed effect of tDCS upon learning; real tDCS was associated with improved learning rates in the WM and PL task, but only after consolidation.

Knowledge transfer for efficient cross domain ranking using AdaRank algorithm

Learning-to-rank has been an exciting topic of research exclusively in hypothetical and the productions in the information retrieval practices. Usually, in the learning-based ranking procedures, it is expected the training and testing data are recovered from the identical data delivery. However those existing research methods do not work well in case of multiple documents retrieved from the cross domains (different domains). In this case ranking of documents would be more difficult where the contents are described in multiple documents from different cross domains. The main goal of this research method is to rank the documents gathered from the multiple domains with improved learning rate by learning features from different domains. The feature level information allocation and instance level information relocation are achieved with four learners namely RankNet, ranking support vector machine (SVM), RankBoost and AdaRank. The estimation results presented that the AdaRank algorithm achieves good performance.

Innovative Target Tracking Method Combined Adaptive $\alpha$ -$\beta$ Filter With Robust BPNN

Target tracking is popular in computer vision field. Although the classic BPNN completes targets tracking, its computation is complex and tracking accuracy is low when the tracking scene is uncertain or complex. To deal with the difficulties above, in this paper, we propose an innovative target tracking method combined adaptive α-β filter with robust BPNN. First, we utilize the adaptive α-β filter to compute the location region on optimal filtering parameters in the prediction stage. Of course, the novel filter reduces the region and gives effective image information to the robust BPNN that has the optimal number and weight of neurons as well as the improved learning rate. Subsequently, the network makes an accurate recognition and sends back the updated positions of targets to the filter for the next cycle. Employing the novel interactive mechanism, the numerical study and experiments indicate that the proposed method has remarkable improvement on average performance in the uncertain and complex environment.

Learning Curve, Change in Industrial Environment, and Dynamics of Production Activities in Unconventional Energy Resources

Since 2007, shale oil and gas production in the United States has become a significant portion of the global fossil fuel market. The main cause for the increase in production of shale oil and gas in the US is the adoption of new production technologies, namely, horizontal drilling and hydraulic fracturing. However, the production cost of shale oil and gas in the US is comparably higher than the production cost of conventional oil and gas. In 2014, the crude oil and natural gas price decreased significantly to approximately 40 dollars per barrel, and natural gas prices decreased to 3 dollars per million British thermal unit, and thus the productivity and financial conditions for the exploration and production of shale oil and natural gas for producers in the United States have worsened critically. Therefore, technological innovation has become one of the most interesting issues of the energy industry. The present study analyzes the trends in technological innovation having a relationship with production activities. This study calculates the learning rate of 30 companies from the petroleum exploration and production industry in the United States using an improved learning rate calculation formula that reflects the changes in the oil production ratio. Thus, more statistically confident calculation results and interpretations of strategic production activities with regard to changes in the industrial environment were achieved in this study.

Error Analysis on Hérmite Learning with Gradient Data

This paper deals with Hermite learning which aims at obtaining the target function from the samples of function values and the gradient values. Error analysis is conducted for these algorithms by means of approaches from convex analysis in the framework of multi-task vector learning and the improved learning rates are derived.

KNOWLEDGE TRANSFER FOR EFFICIENT CROSS DOMAIN RANKING USING ADARANK ALGORITHM

Learning-to-rank has been an exciting topic of research exclusively in hypothetical and the productions in the information retrieval practices. Usually, in the learning-based ranking procedures, it is expected the training and testing data are recovered from the identical data delivery. However those existing research methods do not work well in case of multiple documents retrieved from the cross domains (different domains). In this case ranking of documents would be more difficult where the contents are described in multiple documents from different cross domains. The main goal of this research method is to rank the documents gathered from the multiple domains with improved learning rate by learning features from different domains. The feature level information allocation and instance level information relocation are achieved with four learners namely RankNet, ranking support vector machine (SVM), RankBoost and AdaRank. The estimation results presented that the AdaRank algorithm achieves good performance.

Combined motor point associative stimulation (MPAS) and transcranial direct current stimulation (tDCS) improves plateaued manual dexterity performance

Motor point associative stimulation (MPAS) in hand muscles is known to modify motor cortex excitability and improve learning rate, but not plateau of performance, in manual dexterity tasks. Central stimulation of motor cortex, such as transcranial direct current stimulation (tDCS), can have similar effects if accompanied by motor practice, which can be difficult and tiring for patients. Here we asked whether adding tDCS to MPAS could improve manual dexterity in healthy individuals who are already performing at their plateau, with no motor practice during stimulation. We hypothesized that MPAS could provide enough coordinated muscle activity to make motor practice unnecessary, and that this combination of stimulation techniques could yield improvements even in subjects at or near their peak. If so, this approach could have a substantial effect on patients with impaired dexterity, who are far from their peak. MPAS was applied for 30min to two right hand muscles important for manual dexterity. tDCS was simultaneously applied over left sensorimotor cortex. The motor cortex input/output (I/O) curve was assessed with transcranial magnetic stimulation (TMS), and manual dexterity was assessed with the Purdue Pegboard Test. Compared to sham or cathodal tDCS combined with MPAS, anodal tDCS combined with MPAS significantly increased the plateau of manual dexterity. This result suggests that MPAS has the potential to substitute for motor practice in mediating a beneficial effect of tDCS on manual dexterity.

Increased Adaptation Rates and Reduction in Trial-by-Trial Variability in Subjects with Cerebral Palsy Following a Multi-session Locomotor Adaptation Training.

Cerebral Palsy (CP) results from an insult to the developing brain and is associated with deficits in locomotor and manual skills and in sensorimotor adaptation. We hypothesized that the poor sensorimotor adaptation in persons with CP is related to their high execution variability and does not reflect a general impairment in adaptation learning. We studied the interaction between performance variability and adaptation deficits using a multi-session locomotor adaptation design in persons with CP. Six adolescents with diplegic CP were exposed, during a period of 15 weeks, to a repeated split-belt treadmill perturbation spread over 30 sessions and were tested again 6 months after the end of training. Compared to age-matched healthy controls, subjects with CP showed poor adaptation and high execution variability in the first exposure to the perturbation. Following training they showed marked reduction in execution variability and an increase in learning rates. The reduction in variability and the improvement in adaptation were highly correlated in the CP group and were retained 6 months after training. Interestingly, despite reducing their variability in the washout phase, subjects with CP did not improve learning rates during washout phases that were introduced only four times during the experiment. Our results suggest that locomotor adaptation in subjects with CP is related to their execution variability. Nevertheless, while variability reduction is generalized to other locomotor contexts, the development of savings requires both reduction in execution variability and multiple exposures to the perturbation.

The improved learning rate for regularized regression with RKBSs

The investigation on the performance of learning from samples of functions in Banach spaces is a new research field. A key theoretical problem we need to investigate is how the convergence rate is influenced by the geometry property of the Banach spaces. In the present paper, we provide the learning rate for the kernel regularized regression based on reproducing kernel Banach spaces. The rate is provided in both expected mean and empirical mean. The results show that the uniform convexity influences the learning rate.

Medial-Frontal Stimulation Enhances Learning in Schizophrenia by Restoring Prediction Error Signaling.

When there is a difference between what we expect to happen and what we actually experience, our brains generate a prediction error signal, so that we can map stimuli to responses and predict outcomes accurately. Theories of schizophrenia implicate abnormal prediction error signaling in the cognitive deficits of the disorder. Here, we combine noninvasive brain stimulation with large-scale electrophysiological recordings to establish a causal link between faulty prediction error signaling and learning deficits in schizophrenia. We show that it is possible to improve learning rate, as well as the neural signature of prediction error signaling, in patients to a level quantitatively indistinguishable from that of healthy subjects. The results provide mechanistic insight into schizophrenia pathophysiology and suggest a future therapy for this condition.