Similar Error Research Articles

Online Synchronous Professional Development to Support Teachers of Students With Visual Impairments in Learning Nemeth Code Within Unified English Braille Contexts

Introduction: This study explored the impact of synchronous professional development regarding Nemeth Code Within Unified English Braille (UEB) Contexts on the knowledge and skills of 17 teachers of students with visual impairments. Methods: Participants attended the professional development sessions in June 2022 and submitted a pretest, posttest, final assignment, and posttraining survey. Results: Sixteen out of 17 participants scored higher on their posttest with a mean difference of 5.53. None of the participants submitted an error-free assignment. Total numbers of errors ranged from 8 to 60, and the categories with the highest percentage of errors were formatting, use of opening and terminating switch indicators, and Nemeth Code symbols. Discussion: In order to successfully progress in the general education curriculum, students who read braille must have access to timely, accurately prepared materials in all content areas including mathematics. Direct examination of the final assignment submitted by the 17 participants revealed similar error patterns to previous research. In addition, some of the participants struggled to correctly use Nemeth Code switch indicators. Implications for Practitioners: Teachers of students with visual impairments should continue to build their skills for Nemeth Code within UEB Contexts. Additional resources and professional development are urgently needed.

Deep generative modeling-based data augmentation with demonstration using the BFBT benchmark void fraction datasets

Deep learning (DL) has achieved remarkable successes in many disciplines such as computer vision and natural language processing due to the availability of “big data”. However, such success cannot be easily replicated in many nuclear engineering problems because of the limited amount of training data, especially when the data comes from high-cost experiments. To overcome such a data scarcity issue, this paper explores the applications of deep generative models (DGMs) that have been widely used for image data generation to scientific data augmentation. DGMs, such as generative adversarial networks (GANs), normalizing flows (NFs), variational autoencoders (VAEs), and conditional VAEs (CVAEs), can be trained to learn the underlying probabilistic distribution of the training dataset. Once trained, they can be used to generate synthetic data that are similar to the training data and significantly expand the dataset size. By employing DGMs to augment TRACE simulated data of the steady-state void fractions based on the NUPEC Boiling Water Reactor Full-size Fine-mesh Bundle Test (BFBT) benchmark, this study demonstrates that VAEs, CVAEs, and GANs have comparable generative performance with similar errors in the synthetic data, with CVAEs achieving the smallest errors. The findings shows that DGMs have a great potential to augment scientific data in nuclear engineering, which proves effective for expanding the training dataset and enabling other DL models to be trained more accurately.

The Application of Teacher-Student Collaborative Assessment (TSCA) in Private Colleges

College students’ writing abilities were slowly progressing in the Chinese context. This problem was mainly due to the over-dependence on the teacher’s feedback and the confusion about the writing assessment criteria. This resulted in the low ability of students to self-evaluate their essays. Thus, teacher-student collaborative assessment (TSCA) was introduced to guide students in finding the problems that existed in their peers’ writing which in reverse could also help them detect similar errors in their writing. This study aimed to explore the practicality of the TSCA principle in private colleges with students’ English proficiency below the average. Altogether 42 sophomores majoring in English were selected as the participants. They had taken English writing courses for one year and had some foundation. The questionnaire was given before the course to acquire students’ attitudes towards the TSCA principle. Then students’ writing scores for four assignments were collected during the implementation of the TSCA principle. The results showed that it was applicable to implement the TSCA principle in private colleges for lower-proficient students with their high willingness to this principle. Hopefully, students could have a general idea of writing assessment criteria and do self-evaluation in later autonomous learning.

Taxi fare prediction based on multiple machine learning models

The use of taxis as a fundamental mode of transportation in everyday life has led to the increased popularity of various ride-hailing applications such as Uber and Lyft, enabling users to conveniently request and view the predicted fare for their desired destination. Accurately predicting the fare is thus of significant importance. In this study, machine learning models were employed to forecast taxi fares based on factors such as distance and passenger count. As the initial data only contained latitude and longitude values, the Haversine formula was utilized to calculate the distance between two locations. Moreover, the raw data was plagued with inconsistencies such as negative fares and grossly exaggerated distances, which were resolved by implementing four data cleaning criteria. Following the preprocessing stage, three distinct models (i.e., linear regression, decision tree, and random forest) were trained and evaluated using the root mean square error metric. The results indicated that the random forest model produced the smallest error (1.264), followed by the decision tree model with a similar error rate (1.277), and lastly, the linear regression model with the highest error (1.718). Thus, the random forest model demonstrated superior performance and is recommended for accurate fare predictions.

Hydrodynamics Analysis of Mixing Process in Fermenter

The perfect mixing process ensures the ideal substrate conditions for microorganisms to live to produce ethanol in the fermenter. The critical factors that affect the mixing process are the agitation speed, baffle configuration, and the impeller type. The Computational Fluid Dynamics (CFD) software can make the study of fermenter hydrodynamics more convenient due to the cost savings of expensive apparatus and less time-consuming. This study aimed to examine the hydrodynamic characteristics of a fermenter using the ANSYS FLUENT 2021 R1 software and its validation. The hydrodynamics of the mixing process in the dual-pitched blade impeller bioethanol fermenter in agitation speeds of 200 to 1000 RPM and the effect of baffles were observed by simulation and experimentally. The realizable k-epsilon turbulence model and the Eulerian volume of fluid mixture multiphase model were used in this CFD simulation. The simulation results were relatively close to the experimental with similar flow patterns and low mixing time error, which is 7.9% on average. The results show that the higher the agitation speed, the higher the torque, power, and shear stress. The increasing agitation speed caused the lower mixing time. A faster mixing time was obtained in the fermenter with no baffle configuration.

Verification of different Fizeau fringe analysis algorithms based on airborne wind lidar data in support of ESA's Aeolus mission.

The Aeolus mission by the European Space Agency was launched in August 2018 and stopped operations in April 2023. Aeolus carried the direct-detection Atmospheric LAser Doppler INstrument (ALADIN). To support the preparation of Aeolus, the ALADIN Airborne Demonstrator (A2D) instrument was developed and applied in several field campaigns. Both ALADIN and A2D consist of so-called Rayleigh and Mie channels used to measure wind from both molecular and particulate backscatter signals. The Mie channel is based on the fringe-imaging technique, which relies on determining the spatial location of a linear interference pattern (fringe) that originated from multiple interference in a Fizeau spectrometer. The accuracy of the retrieved winds is among others depending on the analytic algorithm used for determining the fringe location on the detector. In this paper, the performance of two algorithms using Lorentzian and Voigt fit functions is investigated by applying them to A2D data that were acquired during the AVATAR-I airborne campaign. For performance validation, the data of a highly accurate heterodyne detection wind lidar (2-µm DWL) that was flown in parallel are used as a reference. In addition, a fast and non-fit-based algorithm based on a four-pixel intensity ratio approach (R 4) is developed. It is revealed that the Voigt-fit-based algorithm provides 50% more data points than the Lorentzian-based algorithm while applying a quality control that yields a similar random error of about 1.5m/s. The R 4 algorithm is shown to deliver a similar accuracy as the Voigt-fit-based algorithms, with the advantage of a one to two orders of magnitude faster computation time. Principally, the R 4 algorithm can be adapted to other spectroscopic applications where sub-pixel knowledge of the location of measured peak profiles is needed.

Accidental lipid overdose in a preterm infant: A case report.

Intravenous lipid emulsions (ILEs) provide essential fatty acids during parenteral nutrition (PN). Serious adverse events including death can occur from overdose. We report an accidental overdose in a preterm infant. On Day 2 of life, a 29-week gestational age (GA) twin was accidentally given 47.5 mL of Intralipid20% (≈3x daily amount) in 50-minutes. No apparent clinical deterioration occurred, although blood samples were lipaemic. Outcomes at 2 years corrected GA were similar to that of his twin. Service changes were made to infusion packaging and administration to avoid similar errors. Medication errors in neonates are unfortunately common. Published articles usually focus on poor outcomes, which can increase the distress for parents of children where errors have occurred. Publishing the full spectrum of outcomes instead allows parents and professionals to be aware of all possibilities and lessons learnt, even if serious harm was avoided.

Effects of Optimal Tooth Microgeometry Modifications on Static and Dynamic Transmission Errors of Hybrid Spur Gear Drivetrains

A hybrid gear concept that combines a metallic outer rim of gear teeth with a composite web to reduce drivetrain weight was evaluated for impact of tooth microgeometry modifications on transmission error. Control of transmission error through tooth microgeometry modification is important for control of noise and vibrations generated by a drivetrain. The added flexibility of hybrid gears over steel gears brings to question the performance of hybrid over conventional gears relative to their dynamic transmission error and resulting noise levels. Previously developed drivetrain models featuring hybrid spur gears were used to determine optimal tooth microgeometry modifications that minimized peak-to-peak transmission error. Static and dynamic transmission errors were then calculated using the optimal microgeometries and compared to results for a similarly optimized all-steel drivetrain. From the results, it appears that the use of hybrid gears will not negatively affect vibration performance for low- and medium-speed applications, as hybrid gear models predicted similar transmission errors to their all-steel counterparts. At higher speeds, drivetrains featuring hybrid gears were predicted to have significantly different transmission errors, but whether this difference was an improvement or detriment is design and speed-dependent. Therefore, careful design is necessary for high-speed hybrid gears.

Systematic comparison of genome information processing and boundary recognition tools used for genomic island detection

Genomic islands are fragments of foreign DNA that are found in bacterial and archaeal genomes, and are typically associated with symbiosis or pathogenesis. While numerous genomic island detection methods have been proposed, there has been limited evaluation of the efficiency of the genome information processing and boundary recognition tools. In this study, we conducted a review of the statistical methods involved in genomic signatures, host signature extraction, informative signature selection, divergence measures, and boundary detection steps in genomic island prediction. We compared the performances of these methods on simulated experiments using alien fragments obtained from both artificial and real genomes. Our results indicate that among the nine genomic signatures evaluated, genomic signature frequency and full probability performed the best. However, their performance declined when normalized to their expectations and variances, such as Z-score and composition vector. Based on our experiments of the E. coli genome, we found that the confidence intervals of the window variances achieved the best performance in the signature extraction of the host, with the best confidence interval being 1.5–2 times the standard error. Ordered kurtosis was most effective in selecting informative signatures from a single genome, without requiring prior knowledge from other datasets. Among the three divergence measures evaluated, the two-sample t-test was the most successful, and a non-overlapping window with a small eye window (size 2) was best suited for identifying compositionally distinct regions. Finally, the maximum of the Markovian Jensen-Shannon divergence score, in terms of GC-content bias, was found to make boundary detection faster while maintaining a similar error rate.

Flow count data-driven static traffic assignment models through network modularity partitioning

AbstractAccurate static traffic assignment models are important tools for the assessment of strategic transportation policies. In this article we present a novel approach to partition road networks through network modularity to produce data-driven static traffic assignment models from loop detector data on large road systems. The use of partitioning allows the estimation of the key model input of Origin–Destination demand matrices from flow counts alone. Previous network tomography-based demand estimation techniques have been limited by the network size. The amount of partitioning changes the Origin–Destination estimation optimisation problems to different levels of computational difficulty. Different approaches to utilising the partitioning were tested, one which degenerated the road network to the scale of the partitions and others which left the network intact. Applied to a subnetwork of England’s Strategic Road Network and other test networks, our results for the degenerate case showed flow and travel time errors are reasonable with a small amount of degeneration. The results for the non-degenerate cases showed that similar errors in model prediction with lower computation requirements can be obtained when using large partitions compared with the non-partitioned case. This work could be used to improve the effectiveness of national road systems planning and infrastructure models.

A Green Bridge: Enhancing a Multi-Pesticide Test for Food by Phase-Transfer Sample Treatment Coupled with LC/MS.

The preparation and treatment of the sample has become an important part of the determination process, which directly affects the accuracy of detection. The preparation of the sample for final detection is actually a process of separation and transfer of the target to be tested from the sample matrix. The phase-transfer process of analysis and detection is the process of transferring the target substance to be measured from a complex multiphase system to a simple homogeneous system. This study shows a new phase-transfer process for food sample pretreatment in the determination of carbamate pesticides. Edible gum, xanthan gum, carrageenan, and gelatin were selected for purification testing from the perspective of eco-friendliness and safety. Phase-transfer purification process research was carried out on spinach and other foods. Compared with the commonly used QuEChERS method, the LC/MS results indicate that the straightforward carrageenan treatment process can significantly diminish the detection matrix effect and yield similarly superior detection parameters. The phase-transfer purification method with carrageenan has similar sensitivity and systematic error. The limits of detection and limits of quantitation of each pesticide compound in six plant sample substrates were 0.02-0.36 μg/kg and 0.06-1.9 μg/kg, respectively, which were lower than the residue limits here and abroad. Supplemental recoveries in six blank samples at 5, 20, and 100 μg/kg with the phase-transfer process method were better than those for the QuEChERS method. Positive determination results of actual samples using carrageenan phase-transfer purification proved that this method can be used for related detection from a practical point of view.

Concatenated Forward Error Correction With KP4 and Single Parity Check Codes

Concatenated forward error correction is studied using an outer KP4 Reed-Solomon code with hard-decision decoding and inner single parity check (SPC) codes with Chase/Wagner soft-decision decoding. Analytical expressions are derived for the end-to-end frame and bit error rates for transmission over additive white Gaussian noise channels with binary phase-shift keying (BPSK) and quaternary amplitude shift keying (4-ASK), as well as with symbol interleavers and quantized channel outputs. The BPSK error rates are compared to those of two other inner codes: a two-dimensional product code with SPC component codes and an extended Hamming code. Simulation results for unit-memory inter-symbol interference channels and 4-ASK are also presented. The results show that the coding schemes achieve similar error rates, but SPC codes have the lowest complexity and permit flexible rate adaptation.

Scaling-up RADseq methods for large datasets of non-invasive samples: Lessons for library construction and data preprocessing.

Genetic non-invasive sampling (gNIS) is a critical tool for population genetics studies, supporting conservation efforts while imposing minimal impacts on wildlife. However, gNIS often presents variable levels of DNA degradation and non-endogenous contamination, which can incur considerable processing costs. Furthermore, the use of restriction-site-associated DNA sequencing methods (RADseq) for assessing thousands of genetic markers introduces the challenge of obtaining large sets of shared loci with similar coverage across multiple individuals. Here, we present an approach to handling large-scale gNIS-based datasets using data from the spotted hyena population inhabiting the Ngorongoro Crater in Tanzania. We generated 3RADseq data for more than a thousand individuals, mostly from faecal mucus samples collected non-invasively and varying in DNA degradation and contamination level. Using small-scale sequencing, we screened samples for endogenous DNA content, removed highly contaminated samples, confirmed overlap fragment length between libraries, and balanced individual representation in a sequencing pool. We evaluated the impact of (1) DNA degradation and contamination of non-invasive samples, (2) PCR duplicates and (3) different SNP filters on genotype accuracy based on Mendelian error estimated for parent-offspring trio datasets. Our results showed that when balanced for sequencing depth, contaminated samples presented similar genotype error rates to those of non-contaminated samples. We also showed that PCR duplicates and different SNP filters impact genotype accuracy. In summary, we showed the potential of using gNIS for large-scale genetic monitoring based on SNPs and demonstrated how to improve control over library preparation by using a weighted re-pooling strategy that considers the endogenous DNA content.

Bayesian Hyperparameter Optimization of stacked Bidirectional Long Short-Term Memory neural network for the State of Charge estimation

The battery industry has recently grown as a result of electric power adoption in many applications showing less reliance on fossil fuels. Accurate estimation of the State of Charge (SOC) is vital for optimal Battery Management System (BMS) operation to maintain the use of the battery within the safety limits. Data-driven methods require a careful hyperparameter selection as their results strongly depend on their choice. Heuristic methods involve manual tuning or exhaustive ways, which can be computationally demanding. Here a Bayesian Hyperparameter Optimization (BHO) with a Gaussian process is proposed, exploiting the benefits of the probabilistic approach to hyperparameter tuning to reduce the computational effort. For the accurate prediction of SOC, the BHO technique is implemented on a stacked Bidirectional Long Short-Term Memory (BiLSTM) neural network, incorporating a novel set of hyperparameters. The presented method is validated using two well-documented public datasets. Results are compared using datasets with different time granularity, revealing accuracy and computational differences while exhibiting superior yield capabilities compared to other state-of-the-art methods. To assess the computational effort associated with data size and model creation, a time granularity analysis was performed. The analysis demonstrated that decreasing the timestep resulted in reduced network creation time while maintaining similar error levels. Moreover, in order to ensure the comparability of the method, the metric of Floating-point Operations (FLOPs) was integrated to quantify the network’s computational volume. Finally, a sensitivity analysis performed on the time-series window span showed accuracy and computational load differences highlighting the need for caution and careful consideration.

Automated cross-sectional view selection in CT angiography of aortic dissections with uncertainty awareness and retrospective clinical annotations

Surveillance imaging of patients with chronic aortic diseases, such as aneurysms and dissections, relies on obtaining and comparing cross-sectional diameter measurements along the aorta at predefined aortic landmarks, over time. The orientation of the cross-sectional measuring planes at each landmark is currently defined manually by highly trained operators. Centerline-based approaches are unreliable in patients with chronic aortic dissection, because of the asymmetric flow channels, differences in contrast opacification, and presence of mural thrombus, making centerline computations or measurements difficult to generate and reproduce. In this work, we present three alternative approaches – INS, MCDS, MCDbS – based on convolutional neural networks and uncertainty quantification methods to predict the orientation (ϕ,θ) of such cross-sectional planes. For the monitoring of chronic aortic dissections, we show how a dataset of 162 CTA volumes with overall 3273 imperfect manual annotations routinely collected in a clinic can be efficiently used to accomplish this task, despite the presence of non-negligible interoperator variabilities in terms of mean absolute error (MAE) and 95% limits of agreement (LOA). We show how, despite the large limits of agreement in the training data, the trained model provides faster and more reproducible results than either an expert user or a centerline method. The remaining disagreement lies within the variability produced by three independent expert annotators and matches the current state of the art, providing a similar error, but in a fraction of the time.

Accounting for seasonality in the metastatistical extreme value distribution

The typical approach to account for non-stationarity in the generalized extreme value distribution (GEV) is to model the temporal behavior of the GEV parameters, e.g.,with linear relationships between the parameters and e.g.,the year. When in addition, seasonality, i.e., sub-yearly patterns are of interest, this relationship needs to be more complex. A natural candidate for nonstationary analysis of precipitation extremes is the metastatistical extreme value distribution (MEVD) assuming the ordinary rainfall events being Weibull distributed. It uses all ordinary events of the underlying rainfall distribution and implicitly considers temporal nonstationary changes in the Weibull parameters by a blockwise estimation. However, the MEVD still does not make use of the sub-yearly evolution of the distribution parameters and thus discards a significant part of information. In this paper a modification of the MEVD is proposed. The temporal MEVD (TMEV) relaxes the assumption of constant coefficients of the underlying distribution. The MEVD in each block estimates the Weibull parameters using the number of wet events in that block, regardless whether an individual wet day contributes much or less to the yearly rainfall sum. The newly proposed TMEV weights the single wet days contributions during parameter estimation and is therefore able to explicitly account for seasonal differences of rainfall events. It is shown that the TMEV provides a very similar error characteristic to the simplified MEVD for the estimation of quantiles with different sample lengths. In addition, we demonstrate the ability of the TMEV to identify longterm trends and seasonal variations of extreme precipitation in Austria and discuss possible implications for general trend estimations. We also present a spatio-temporal TMEV model, which is able to reproduce known patterns of a 50-year return level map of daily rainfall in Austria, thereby lending credence to the TMEV approach.

Leone Caetani’nin Üsâme b. Zeyd’in Belkâ Seferi Hakkındaki Görüşlerine Dair Tahlil, Tenkit ve Değerlendirme

This paper deals with the views and claims of the Italian politician and orientalist Leone Caetani in his work Annali Dell'Islām, which deals with the Prophet's preparation of an army to be sent to the Syrian region before his death and his appointment of Usāma b. Zayd as the commander of the military, which took place in the 11th year of the Hijrah in Islamic history. Due to its importance, every Muslim wants to learn the Islamic sciences from their basic sources accurately, especially the life of the Prophet Muhammad. For this purpose, Muslim scholars researched this field and produced many works. Western scholars also conducted research in this field for different purposes and eventually produced works within the scope of Orientalism. Leone Caetani (d. 1935) was one of the orientalists who aimed to study the religion, culture, thought, economic, and social life of the East. He wrote his work Annali Dell'Islām in ten volumes. The translation of the work from Italian by Hüseyin Cahit Yalçın between 1924 and 1927 caused a great outrage in our country. Due to the fact that historians live in different geographies and are nourished by the culture in which they live, they have often failed to do so, even though they want to approach events objectively. Therefore, it is important to approach orientalist writers' works carefully and evaluate their content. In addition, when differences in beliefs are added to the aforementioned factors, the picture that emerges can be far from scientific. As a matter of fact, this situation is also observed in Leone Caetani. Caetani's work contains many subjective, negative thoughts and incomplete and erroneous information. Although it is the product of many years and labor, it is seen that the work moves away from objectivity and scientificity at some points. Babanzâde Ahmet Naim, and Yusuf Ziya Yörükân wrote articles against his subjective and negative views which do not reflect the truth, and Âsım Köksal wrote a refutation to the first volume. However, when we searched the literature, we realized that previous studies did not include this subject, and the need for research arose. In this study, the information given by Caetani under the title "Preparations for Usāma's departure to Balqa" is subjected to analysis, criticism and evaluation. The aforementioned subject is discussed within orientalist Leone Caetani's views and claims. For this purpose, the main sources such as al-Wāqidī's al-Mag̲hāzī, Ibn Hishām's al-Sīra al-nabawiyya, al-Ṭabarī's Tarikh al-Umam wa al-Muluk, and many others such as the works of Ibn Saʿd, al-Yaʿqūbī, al-Masʿūdī, Yāqūt al-Ḥamawī, al-hahabī, Ibn Kathīr, etc., and the al-Kutub al-sitta in the field of hadīth have been examined to obtain information on the subject. In the light of the information obtained, Caetani's views and claims were objectively evaluated. As a result of our evaluations, we have found both accurate information and erroneous points. For example, while quoting information from al-Masʿūdī, he expressed Usāma's campaigns in the 10th and 11th years of Hijrī as if they took place in the same time without specifying the difference in years. This and similar errors have been explained with the information we have obtained from the main sources.

Implementation of Performance Assessment with Assistance to Improve Students' Scientific Written Communication Skills

The purpose of this study is to describe how the application of performance appraisal with mentoring and improvement of scientific written communication skills. This study used Classroom Action Research (CAR), which consisted of 2 cycles and was carried out on 35 class VII I SMPN 6 student subjects. The results of this study describe performance appraisal activities with assistance and improvement of scientific written communication skills in general and in all indicators. Thus, it can be concluded that the implementation of performance appraisal with assistance can improve students' scientific written communication skills. Performance assessment is carried out by giving assignments in the form of worksheets (LK), while the assistance steps carried out consist of giving written feedback from the teacher to student worksheets, grouping students based on similar errors, students reading the results of the feedback, the teacher visits the group, debriefing the teacher with students, in groups peer tutors are also carried out, and LK improvement by students. The action step that is considered the most important in influencing the results of the action is the mentoring step, namely grouping students based on similar errors. The increase in scientific written communication that scored ≥ 70 from cycle 1 (63.89%) to cycle 2 (97.14%) was 33.25%. This increase is also supported by an increase in each scientific written communication indicator, namely those who get a score of ≥ 75 for the 1st indicator making a description from a graph or table 33.4%, the 2nd indicator changing the description data into a table 5.6%, the 3rd indicator changing data from tables to graphs is 63.9%, the 4th indicator infers from data in the form of tables or graphs 62.63%. The action step that is considered the most important in influencing the results of the action is the mentoring step, namely grouping students based on similar errors. The increase in scientific written communication that scored ≥ 70 from cycle 1 (63.89%) to cycle 2 (97.14%) was 33.25%. This increase is also supported by an increase in each scientific written communication indicator, namely those who get a score of ≥ 75 for the 1st indicator making a description from a graph or table 33.4%, the 2nd indicator changing the description data into a table 5.6%, the 3rd indicator changing data from tables to graphs is 63.9%, the 4th indicator infers from data in the form of tables or graphs 62.63%. The action step that is considered the most important in influencing the results of the action is the mentoring step, namely grouping students based on similar errors. The increase in scientific written communication that scored ≥ 70 from cycle 1 (63.89%) to cycle 2 (97.14%) was 33.25%. This increase is also supported by an increase in each scientific written communication indicator, namely those who get a score of ≥ 75 for the 1st indicator making a description from a graph or table 33.4%, the 2nd indicator changing the description data into a table 5.6%, the 3rd indicator changing data from tables to graphs is 63.9%, the 4th indicator infers from data in the form of tables or graphs 62.63%.

Neural Architecture Search Based on a Multi-Objective Evolutionary Algorithm With Probability Stack

With the emergence of deep neural networks, many research fields, such as image classification, object detection, speech recognition, natural language processing, machine translation and automatic driving, have made major breakthroughs in technology and the research achievements have been successfully applied in many real-life applications. Combining evolutionary computation and neural architecture search (NAS) is an important approach to improve the performance of deep neural networks. Usually, the related researchers only focus on precision. Thus, the searched neural architectures always perform poorly in the other indexes such as time cost. In this paper, a multi-objective evolutionary algorithm with a probability stack (MOEA-PS) is proposed for NAS, which considers the two objects of precision and time consumption. MOEA-PS uses an adjacency list to represent the internal structure of deep neural networks. Besides, a unique mechanism is introduced into the multi-objective genetic algorithm to guide the process of crossover and mutation when generating offspring. Furthermore, the structure blocks are stacked using a proxy model to generate deep neural networks. The results of the experiments on Cifar-10 and Cifar-100 demonstrate that the proposed algorithm has a similar error rate compared with the most advanced NAS algorithms, but the time cost is lower. Finally, the network structure searched on Cifar-10 is transferred directly to the ImageNet dataset, which can achieve 73.6% classification accuracy.

The development of complex engineering models using artificial neural network-based proxy models for life cycle assessments of energy systems

The energy industry has been using life cycle assessment (LCA) to determine the environmental impact of projects. Obtaining accurate data of certain industrial activities requires complex engineering models that have long computing times, are difficult for non-experts to use, and may contain confidential data. This work examines using proxy models based on quadratic and artificial neural network (ANN) regression to create an accurate, easy-to-use, black-box model that can be easily shared. Generating target values from the engineering software needed for training can be time-consuming, hence, adaptive sampling methods are examined (random, spread, high error, and the combo method [50/50 random/high error]). Two case studies were examined: a transportation fuel LCA of Maya, Bow River, and mined bitumen crude oils; and an LCA of a natural gas transmission pipeline (NGTL). This work found that ANN proxy models are more accurate than quadratic regression, and the high error sampling method reduced the maximum error but increased the average error. The combo and high error methods using 3000 to 4000 samples achieved similar maximum errors to the random method using 10,000 samples. Because of uncertainty in LCA input values, reducing average error is less valuable than reducing extreme errors. For the NGTL case study, the ANN model was able to reduce the average and max error by 67% and 68%, respectively, while using 35% fewer coefficients; ANN models are more appropriate for complex nonlinear models.