Increase Training Efficiency Research Articles

Towards automatic farrowing monitoring-A Noisy Student approach for improving detection performance of newborn piglets.

Nowadays, video monitoring of farrowing and automatic video evaluation using Deep Learning have become increasingly important in farm animal science research and open up new possibilities for addressing specific research questions like the determination of husbandry relevant indicators. A robust detection performance of newborn piglets is essential for reliably monitoring the farrowing process and to access important information about the welfare status of the sow and piglets. Although object detection algorithms are increasingly being used in various scenarios in the field of livestock farming, their usability for detecting newborn piglets has so far been limited. Challenges such as frequent animal occlusions, high overlapping rates or strong heterogeneous animal postures increase the complexity and place new demands on the detection model. Typically, new data is manually annotated to improve model performance, but the annotation effort is expensive and time-consuming. To address this problem, we propose a Noisy Student approach to automatically generate annotation information and train an improved piglet detection model. By using a teacher-student model relationship we transform the image structure and generate pseudo-labels for the object classes piglet and tail. As a result, we improve the initial detection performance of the teacher model from 0.561, 0.838, 0.672 to 0.901, 0.944, 0.922 for the performance metrics Recall, Precision and F1-score, respectively. The results of this study can be used in two ways. Firstly, the results contribute directly to the improvement of piglet detection in the context of birth monitoring systems and the evaluation of the farrowing progress. Secondly, the approach presented can be transferred to other research questions and species, thereby reducing the problem of cost-intensive annotation processes and increase training efficiency. In addition, we provide a unique dataset for the detection and evaluation of newborn piglets and sow body parts to support researchers in the task of monitoring the farrowing process.

Ecological and Cost Advantage from the Implementation of Flight Simulation Training Devices for Pilot Training

The paper discusses a case study of obtaining an airline pilot license in integrated training—the so-called “from zero to Airline Transport Pilot License”. The environmental implications of simulator-based training were examined across multiple dimensions. Key areas of research include the reduction of harmful exhaust gases pollution associated with traditional flight training activities. Based on our analysis, it can be stated that increasing the use of Flight Simulation Training Devices in pilot training should be significant consideration. This approach brings many benefits, especially ecological ones. Changing the training program and increasing the use of flight simulators can result in a reduction of CO2 emissions by up to 70%. Based on country specific electricity factors, CO2 emissions during flight training in each EU country were calculated. Using Levelized Cost of Electricity average value to calculate training costs in EU countries depends on the mix of energy sources (wind, photovoltaics, carbon and gas). The findings highlight the significant ecological advantages of simulator-based training methods in mitigating the environmental footprint of aviation operations. By seeking to minimize environmental disruption and increase training efficiency, the adoption of simulators is a sustainable approach to pilot training that is consistent with global efforts to mitigate climate change and protect natural ecosystems.

Asynchronous Federated Learning with Grey Wolf Optimization for the Heterogeneity IoT Devices

The Internet of Things (IoT) creates new options for real-time data collection and Machine Learning (ML) model development. Nonetheless, it's feasible that a particular IoT device does not have sufficient computational power to train and implement a complete learning model. However, there are significant communication costs and data security and privacy concerns associated with sending actual data to a centralised server with a lot of computational power. Federated Learning (FL) is a potential way to train ML models using low-powered devices and Edge Servers (ES) since it is a distributed ML architecture. However, the vast majority of the works in existence make the unsustainable assumption of a synchronized parameters update manner with similar IoT nodes and reliable communications networks. To increase training efficiency and accelerate the speed for heterogeneous IoT devices in an unreliable network environment, we designed an Asynchronous Federated Learning strategy with Grey Wolf Optimization (AFL-GWO) in this research. In particular, we develop a Lightweight Node Selection (LNS) technique and propose an AFL-GWO model to efficiently complete learning tasks. To ensure that diverse IoT nodes with varying computational capabilities and network connectivity are represented in the global learning aggregate, the proposed technique makes node selections on an iterative basis. We show through extensive experiments that our suggested AFL-GWO system outperforms the state-of-the-art techniques on identically and independently distributed (IID) and non-IID data distribution in a variety of contexts.

The effect of musk incense stick aroma inhalation on different features of electroencephalogram signals and working memory for use in neurofeedback training

Susceptibility of brain activity to the inhalation of different aromas established the idea of inhaling an aroma as an auxiliary factor in neurofeedback training. The present study has focused on investigating the effect of musk incense stick aroma inhalation on various features of EEG signals (power of different frequency bands along with Katz fractal dimension (Katz’s FD), approximate entropy (ApEn) and scaling exponent) and working memory for use in neurofeedback training. To raise this aim, the EEG signals were recorded under eyes-open (EO) and eyes-closed (EC) conditions for 22 healthy subjects before and during the musk incense stick aroma inhalation, followed by performing Wechsler Memory Scale (WMS). The results showed that musk incense stick aroma inhalation under EO condition increased the relative power of delta, theta, sensorimotor rhythm (SMR) and low beta frequency bands along with ApEn and Katz’s FD in most channels, whereas decreased the relative power of alpha, high beta and gamma bands. A decrease was found in the relative power of delta, theta, SMR, low beta, high beta and gamma frequency bands in some channels under EC conditions. ApEn and Katz’s FD values also increased in most channels under EC conditions following musk incense stick aroma inhalation. The scaling exponent also decreased under the same conditions in most channels. However, these changes were not significant in all channels. The WMS score also increased after inhaling aroma, but not significantly. The findings of this research suggest the use of musk incense stick aroma inhalation during neurofeedback training to increase training efficiency.

Data Integrity Attack in Dynamic State Estimation of Smart Grid: Attack Model and Countermeasures

A smart grid integrates advanced sensors, efficient measurement methods, progressive control technologies, and other techniques and devices to achieve safe, efficient and economical operation of the grid system. However, the diversified and open environment of a smart grid makes energy and information of the smart grid vulnerable to malicious attacks. As a representative cyber-physical attack, the data integrity attack has an extremely severe impact on the grid operation for it can bypass the traditional detection mechanisms by adjusting the attack vector. In this paper, we first present the attack strategy against dynamic state estimation of power grid in the perspective of adversary and formulate the data integrity attack detection problem that has the characteristic of sequential decision making as a partially observable Markov decision process. Then, a deep reinforcement learning-based approach is proposed to detect against data integrity attacks, which utilizes the Long Short-Term Memory layer to extract the state features of previous time steps in determining whether the system is currently under attack. Moreover, the noisy networks are employed to ensure effective agent exploration, which prevents the agent from sticking to the non-optimal policy. The principle of a multi-step learning is adopted to increase the estimation accuracy of Q value. To address the sparse rewards problem, the prioritized experience replay is proposed to increase training efficiency. Simulation results demonstrated that the proposed detection approach surpasses the benchmarks in the comparison metrics: delay error rate and false rate. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —In this paper, we present a deep reinforcement learning-based algorithm to defend against the data integrity attacks of smart grid. Most of the previous works discretized the system states and utilized the current state information to identify whether the system is under attack. For this reason, the detection policy may totally ignored the continuously changing characteristics of the grid states, which will lead to poor detection performance. Moreover, the attacked system states only accounts for a small part of the entire grid operation states, the probability of sampling the experience containing the attack state is extremely small, which limits the learning efficiency of previous RL-based detection approaches. In order to increase the accuracy of detection, we first present the attack strategy against power grid’s dynamic state estimation in the perspective of adversary and formulate the partially observable Markov decision process model of attack detection problem. Moreover, we propose a deep reinforcement learning-based detection approach combining the LSTM network to extract the system state features of the previous time steps to determine whether the system is currently being attacked. To address the sparse rewards problem, the prioritized experience replay is used to increase learning efficiency. The experiments demonstrate the effectiveness of proposed detection scheme compared with benchmarks in terms of detection delay as well as accuracy. In conclusion, the proposed detection scheme is helpful in defending against the data integrity attacks without obtaining the opponent’s strategy in advance and can be conveniently applied to the real-world security management system of smart grid.

OPTIMIZE WAV2VEC2S ARCHITECTURE FOR SMALL TRAINING SET THROUGH ANALYZING ITS PRE-TRAINED MODELS ATTENTION PATTERN.

Transformer-based automatic speech recognition (ASR) systems have shown their success in the presence of large datasets. But, in medical research, we have to create ASR for the non-typical population, i.e. pre-school children with speech disorders, with small training dataset. To increase training efficiency on small datasets, we optimize the architecture of Wav2Vec 2.0, a variation of Transformer, through analyzing its pre-trained model's block-level attention pattern. We show that block-level patterns can serve as an indicator for narrowing down the optimization direction. To ensure the reproducibility of our experiments, we leverage Librispeech-100-clean as training data to simulate the limited data condition. We leverage two techniques, local attention mechanism and cross-block parameter sharing, with counter-intuitive configurations. Our optimized architecture outperforms the vanilla architecture about 1.8% absolute word error rate (WER) on dev-clean and 1.4% on test-clean.

A deep learning image segmentation model for agricultural irrigation system classification

Effective water management requires a large-scale understanding of agricultural irrigation systems and how they shift in response to various stressors. Here, we leveraged advances in Machine Learning and availability of very high resolution remote sensing imagery to help resolve this long-standing issue. To this end, we developed a deep learning model to classify irrigation systems at a regional scale using remote sensing imagery. After testing different model architectures, hyper parameters, class weights and image sizes, we selected a U-Net architecture with a Resnet-34 backbone for this purpose. We applied transfer learning to increase training efficiency and model performance. We considered four irrigation systems as well as urban and background areas as land use/cover classes, and applied the model to 8,600 very high resolution (1 m) images, labeled with ground-truth observations of irrigation types, in a case study in Idaho, USA. Images were obtained from the US Department of Agriculture’s National Agriculture Imagery Program. Our model achieved state-of-the-art performance for segmentation of different classes on the train data (85% to 94%), validation data (72% to 86%), and test data (70% to 86%), which attests to the efficacy of the model for the segmentation of images based on spatial features. Aside from leveraging deep learning and remote sensing for resolving the standing real-world problem of multiple irrigation type segmentation, this study develops and publicly shares labeled data, as well as a trained deep learning model, for irrigation type segmentation that can be applied/transferred to other regions globally. Furthermore, this study offers novel information about the impacts of transfer learning, imbalanced training data, and efficacy of various model structures for multiple irrigation type segmentation.

Design of Augmented Reality Training Content for Railway Vehicle Maintenance Focusing on the Axle-Mounted Disc Brake System

Light maintenance training for electric multiple-unit components of the organization of railway operations is generally conducted using maintenance manuals and work videos, following the guidelines of each organization. These manuals are in the form of booklets, complicated and inconvenient for maintenance operators to carry. Therefore, training content that visualizes maintenance procedures in a three-dimensions (3D) space is necessary to overcome the drawbacks of booklet-type training. In this study, we developed augmented reality (AR)-based training content for railway vehicle maintenance to increase training efficiency. Providing warning signs for risky procedures reduces human error, and transparency control makes trainees check the product hierarchy. A virtual experience based on the maintenance manual is provided to improve maintenance proficiency. An axle-mounted disc brake system maintenance manual is implemented in AR to reflect the requirements of maintenance operators. The convenience of this tool is improved by loading the AR content on a mobile device. Two methods of verification were used: the system usability scale (SUS) survey and training efficiency evaluation. The resulting SUS grade was B (excellent), and the training efficiency improved by 34%.

Predicting the success rate of healthy participants in beta neurofeedback: Determining the factors affecting the success rate of individuals

Despite the considerable success of neurofeedback techniques in the treatment of various neurological disorders and the improvement of cognitive performance of healthy individuals, some people fail to learn how to control their brain activities using neurofeedback. Given the time-consuming and costly nature of neurofeedback, the prediction of people’s success rate in training by neurofeedback is of paramount importance. Therefore, the present study aimed to determine the factors affecting the success rate of 7 healthy women over 10 sessions (30 trials) in terms of enhancement of low beta band activities (βL). The relative power of different frequency bands (delta, theta, alpha and beta) of EEG signals obtained in the first training session was considered as the predictor variable along with the participants’ IQ test score. Afterwards, we assessed the predictor variables’ impact on the mean low beta power (15–18 Hz) values of the participants’ EEG signals in the last session (βL(last sess)). According to the results, the mean low beta power in the first session (βL(sess1)) had the most effect on βL(last sess) (R2 = 73.9 %). In the next stage, we designed three systems using the RBF network, which predicted the βL(last sess), mean score of each participant in the last training session and the slope of βL changes of each subject during the training sessions using βL(sess1) (prediction error < 10−11). The designed prediction system may be able to increase training efficiency with neurofeedback and save time and financial resources.

Teaching Appropriate Feedback Reception Skills Using Computer-based Instruction: A Systematic Replication

ABSTRACT Feedback is a common intervention used to address performance issues in clinical and organizational settings. Most research on feedback focuses on manipulating either the parameters of feedback delivery or the behavior of the feedback provider. Because feedback requires a provider and a recipient, it is likely that the behavior of the feedback recipient also influences the impact of performance feedback. The current study investigated the efficacy of training individuals to receive feedback in an appropriate manner using computer-based instruction (CBI). All three participants demonstrated increases in accuracy of appropriate feedback behaviors following CBI. Two participants also demonstrated slight increases in performance on tasks related to workplace performance. This study extends the application of CBI to a new and complex skill repertoire and examines its possible impact on work performance. We discuss how CBI may increase training efficiency in settings with large numbers of staff who need to be trained in certain skills as compared to traditional in-person training formats.

A(DP) 2SGD: Asynchronous Decentralized Parallel Stochastic Gradient Descent With Differential Privacy.

As deep learning models are usually massive and complex, distributed learning is essential for increasing training efficiency. Moreover, in many real-world application scenarios like healthcare, distributed learning can also keep the data local and protect privacy. Recently, the asynchronous decentralized parallel stochastic gradient descent (ADPSGD) algorithm has been proposed and demonstrated to be an efficient and practical strategy where there is no central server, so that each computing node only communicates with its neighbors. Although no raw data will be transmitted across different local nodes, there is still a risk of information leak during the communication process for malicious participants to make attacks. In this paper, we present a differentially private version of asynchronous decentralized parallel SGD framework, or A(DP) 2SGD for short, which maintains communication efficiency of ADPSGD and prevents the inference from malicious participants. Specifically, Rényi differential privacy is used to provide tighter privacy analysis for our composite Gaussian mechanisms while the convergence rate is consistent with the non-private version. Theoretical analysis shows A(DP) 2SGD also converges at the optimal O(1/√T) rate as SGD. Empirically, A(DP) 2SGD achieves comparable model accuracy as the differentially private version of Synchronous SGD (SSGD) but runs much faster than SSGD in heterogeneous computing environments.

Interaction effects of different orders of resistance exercises and rest intervals on performances by young athletes

ObjectiveThe purpose of this study was to verify the acute effects of different exercise orders and rest intervals between sets on young athletes performance. MethodSixteen young male football players (73.2 ± 4.8 kg, 177.5 ± 5.1 cm, BMI 23.2 ± 1.1, 19.8 ± 0.9 years) completed six experimental strength training (ST) sessions with different exercise order sequences (A and B) and rest interval lengths (1 min, 3 min, and self-selected).In Sequence A the exercises were: bench press (BP), back squat (BS), biceps curl (BC) and plantar flexion (PF); while Sequence B was performed in the opposite order (i.e. PF, BC, BS and BP).The total work volume (TWV) per exercise (sets x repetitions x load) and per training session (sum of the TWV of all exercises) were evaluated for all ST sessions. ResultsBC and PF exercises presented higher TWV in sequence B (p ≤ 0.05). Already, the exercises BP, BS and PF presented higher TWV with 3 min and self-selected rest intervals (p ≤ 0.05). The 3-min and self-selected rest intervals presented higher values of TWV per training session compared to the 1-min rest interval (p ≤ 0.05). ConclusionThe exercise order influenced certain exercises (BC and PF), which presented higher TWV when positioned at the beginning of a sequence. While the longer rest intervals (3 min and self-selected) resulted in higher TWV per exercise (BP, BS and PF) and per training session. These results suggest that self-selected rest interval can be implemented to increase training efficiency in young athletes.

Autonomous cage-side system for remote training of non-human primates

BackgroundTraining non-human primates (NHPs) for translational medical experimentation is an essential yet time consuming process. To increase training efficiency, some training systems have been designed for NHPs to use at their home cages. Several autonomous cage-side tablet-based systems have been proposed, but none of these systems allow for remote monitoring and task modification while also being wireless, low-cost, light weight, and portable. New methodHere we present ACTS: an Autonomous Cage-side Training System which meets all these criteria. ACTS consists of 1) a touchscreen tablet and a speaker attached to the subject’s home cage, 2) an inexpensive reward system made from a slightly modified fish feeder, and 3), a laptop operating the system wirelessly and remotely via a router. ResultsWe were able to test the system and wirelessly train two macaques in their home cages. Remote access enabled us to control ACTS from up to 90 m, through up to 3 walls, and through a floor of a building. The device is compatible with different reward pellet sizes and could run about two hours with a ∼4 mm pellet size. The animals were able to generalize the task when transferred to a traditional experimental rig. Comparison with existing methodsThe low cost and modest skill required to build and implement ACTS lowers the barrier for NHP researchers and caregivers to deploy autonomous, remotely controlled tablet-based cage-side systems. ConclusionACTS can be used for low-cost, wireless cage-side training of NHPs being prepared for translational medical experimentation.

Basic Laparoscopic Skills Training Is Equally Effective Using 2D Compared to 3D Visualization: A Randomized Controlled Trial.

Reduced depth perception due to two-dimensional (2D) visualization of a three-dimensional (3D) space represents a main challenge in acquiring basic laparoscopic skills (BLS); 3D visualization might increase training efficiency. This study aimed to assess whether BLS training on a standard box trainer using 2D is at least equally effective compared to 3D. Medical students were randomized to training of Fundamentals of Laparoscopic Surgery (FLS) tasks using either 2D or 3D for four weeks. Baseline and post-training tests were performed using the assigned visualization modality. Data of 31 participants were analyzed (n = 16 2D, n = 15 3D). Baseline test scores did not differ significantly between groups; only at the peg transfer task and total scores, the 3D group performed better than the 2D group. All scores improved significantly in both groups, with post training scores not differing significantly between groups. Non-inferiority of 2D compared to 3D was demonstrated for total score improvement and improvement in all individual FLS tasks except for suturing with extracorporeal knot tying. Post training test performance did not change significantly when changing to the unfamiliar modality. In conclusion, BLS training using standard 2D is at least equally effective as with 3D, without significant disadvantages when changing to the other modality.

Применение программных средств универсального тренажерного комплекса для повышения эффективности обучения персонала химических и нефтехимических производств

Применение программных средств универсального тренажерного комплекса для повышения эффективности обучения персонала химических и нефтехимических производств

Active learning for accuracy enhancement of semantic segmentation with CNN-corrected label curations: Evaluation on kidney segmentation in abdominal CT

Segmentation is fundamental to medical image analysis. Recent advances in fully convolutional networks has enabled automatic segmentation; however, high labeling efforts and difficulty in acquiring sufficient and high-quality training data is still a challenge. In this study, a cascaded 3D U-Net with active learning to increase training efficiency with exceedingly limited data and reduce labeling efforts is proposed. Abdominal computed tomography images of 50 kidneys were used for training. In stage I, 20 kidneys with renal cell carcinoma and four substructures were used for training by manually labelling ground truths. In stage II, 20 kidneys from the previous stage and 20 newly added kidneys were used with convolutional neural net (CNN)-corrected labelling for the newly added data. Similarly, in stage III, 50 kidneys were used. The Dice similarity coefficient was increased with the completion of each stage, and shows superior performance when compared with a recent segmentation network based on 3D U-Net. The labeling time for CNN-corrected segmentation was reduced by more than half compared to that in manual segmentation. Active learning was therefore concluded to be capable of reducing labeling efforts through CNN-corrected segmentation and increase training efficiency by iterative learning with limited data.

Using Virtual Reality with Integrated Eye Tracking for Visual Rehabilitation

A stroke affecting the visual areas of the brain can often cause substantial blind areas in the visual field, called cortical blindness (CB). These blind spots are not caused by damage to the eye itself, but by damage to the visual processing mechanisms in the brain. Although rehabilitation programs have been developed to help CB patients recover visual function, they are limited in the efficiency of the recovery. The first generation of recovery techniques are performed while seated at a desktop computer with the head immobilized in a chin rest. Eye tracking, which is essential to the successful completion of the training, is only done at the first training session supervised by an expert, and the absence of in-home eye tracking during the remainder of the training leads to many wasted trials. Additionally, the training target is very simple, and does not fully stimulate the regions of the brain that enable the recovery. Our aim is to improve upon previously demonstrated training efficiency by leveraging emerging virtual reality (VR) technology with integrated eye tracking. Our new VR-based system is cost effective, can be deployed at home, and facilitates more natural, head-free behavior. The proposed work is a first step towards the long-term goal of developing a more effective at-home rehabilitation paradigm that provides richer a richer stimulus, is more enjoyable, and can increase training efficiency, thus improving patient recovery of visual function.

Overcoming Limitations of the HoloLens for Use in Product Assembly

Augmented Reality (AR) is an emerging technology that could greatly increase training efficiency and effectiveness for assembly processes. In fact, studies show that AR may reduce time and errors by as much as 50%. While many devices are available to display AR content, the use of Head Mounted Displays (HMDs) is the most interesting for manufacturing and assembly as they free up the user&#x2019;s hands. Due to the emerging nature of this technology, there are many limitations including input, field of view, tracking and occlusion. The work presented in this paper explores the use of the Microsoft HoloLens to deliver AR work instructions for product assembly. An AR assembly application was developed to guide a trainee through the assembly of a mock aircraft wing. To ensure accurate instructions were displayed, novel techniques were developed to mitigate the HoloLens&#x2019; tracking and display limitations. Additionally, a data visualization application was developed to validate the training session and explore trends through data collected from the HoloLens along with wearable physiological sensor data.

Modeling of the Three-way Catalytic Converter by Recurrent Neural Networks

For offline development and calibration of On-Board Diagnosis functions of the exhaust gas aftertreatment system a model of the three-way catalytic converter is essential. The goal of the method presented here is to approximate the conversion behavior of a three way catalytic converter by a model which is suitable for application in such a simulation environment. To that purpose we followed a Black Box modeling approach using Recurrent Neural Networks (RNNs) and compared different training and model settings in a case study. In order to increase training efficiency and model accuracy we investigated transfer learning techniques to transfer knowledge from physical models to the RNN model.

A Systematic Review of the Professional Competencies of Coaches in Elite Sport

Expert coach encourages a continuity of the educational process by developing athlete‘s psychomotor development, during attainment of knowledge, techniques, the perfection of tactics; coach accumulates the experience of competitive practice. In order to increase training efficiency, coaches must continually develop existing competencies and acquire new ones. The aim of the study was to reveal the basic competencies of coaching career preparing high performance athletes. Methods of the research: systematic review of documents and scientific literature which regulate sports coaches’ activity in elite sport. [...].