A Survey on Uncertainty Estimation in Deep Learning Classification Systems from a Bayesian Perspective

Decision-making based on machine learning systems, especially when this decision-making can affect human lives, is a subject of maximum interest in the Machine Learning community. It is, therefore, necessary to equip these systems with a means of estimating uncertainty in the predictions they emit in order to help practitioners make more informed decisions. In the present work, we introduce the topic of uncertainty estimation, and we analyze the peculiarities of such estimation when applied to classification systems. We analyze different methods that have been designed to provide classification systems based on deep learning with mechanisms for measuring the uncertainty of their predictions. We will take a look at how this uncertainty can be modeled and measured using different approaches, as well as practical considerations of different applications of uncertainty. Moreover, we review some of the properties that should be borne in mind when developing such metrics. All in all, the present survey aims at providing a pragmatic overview of the estimation of uncertainty in classification systems that can be very useful for both academic research and deep learning practitioners.

Performance and Power Estimation of STT-MRAM Main Memory with Reliable System-level Simulation

It is questionable whether DRAM will continue to scale and will meet the needs of next-generation systems. Therefore, significant effort is invested in research and development of novel memory technologies. One of the candidates for next-generation memory is Spin-Transfer Torque Magnetic Random Access Memory (STT-MRAM). STT-MRAM is an emerging non-volatile memory with a lot of potential that could be exploited for various requirements of different computing systems. Being a novel technology, STT-MRAM devices are already approaching DRAM in terms of capacity, frequency, and device size. Although STT-MRAM technology got significant attention of various major memory manufacturers, academic research of STT-MRAM main memory remains marginal. This is mainly due to the unavailability of publicly available detailed timing and current parameters of this novel technology, which are required to perform a reliable main memory simulation on performance and power estimation. This study demonstrates an approach to perform a cycle accurate simulation of STT-MRAM main memory, being the first to release detailed timing and current parameters of this technology from academia—essentially enabling researchers to conduct reliable system-level simulation of STT-MRAM using widely accepted existing simulation infrastructure. The results show a fairly narrow overall performance deviation in response to significant variations in key timing parameters, and the power consumption experiments identify the key power component that is mostly affected with STT-MRAM.

Child Abuse in the Canadian and American Olympic Movement: Reforming Institutional Oversight of Amateur Sports

For many years, sexual abuse of young athletes quietly festered throughout amateur sport organizations in Canada and the United States. However, the veil has recently been lifted by the highly publicized testimonies of athletes sexually abused as minors by disgraced former USA Gymnastics (“USAG”) physician and Michigan State University professor Larry Nassar. The troubling details of these stories spurred investigations seeking to identify the causes behind the institutional failures to intervene in Nassar’s perpetration of abuse. The evidence gathered by these investigations alongside independent academic research reveal that unfettered predatory behaviour is a pervasive issue across the institutions responsible for overseeing the American Olympic movement and amateur sports. In addition, similar patterns of unchecked abuse have since been identified as plaguing Canadian Olympic and amateur sport organizations. This paper examines the institutional failures to address child sexual abuse occurring under the oversight of Olympic and amateur sport organizations in Canada and the United States. In both countries, efforts are currently underway to reform governance of these institutions to better protect minors participating in amateur sports. Accordingly, this paper also analyzes the policies implemented thus far and makes substantive recommendations on ideal federal level initiatives.

The Impact of Telegram on Syntactic Complexity of the Saudi Female EFL Undergraduates’ Argumentative, Classification, and Reaction Essays

The aim of this exploratory experimental quantitative is to investigate the impact of the learning environment and text types on the syntactic complexity of female Saudi students majoring in English language. Specifically, the study seeks to explore how and when the syntactic complexity of these students increases/decreases as a whole and across the fourteen measures of syntactic complexity (MLS, MLT.MLC.C/S, VP/T, C/T, DC/C, DC/T, T/S, CT/T, CP/T, CP/C, CN/T, CN/C) in two learning contexts: traditional learning context (TLC) and blended learning context (BLT) across three writing tasks (argumentative, classification ,and reaction). It purports, also, to find out when and which learning context leads to the most/least increase/decrease in the syntactic complexity (as a whole and across specific measures) of Saudi undergraduate in the three writing tasks. To answer such questions, 48 female Saudi EFL undergraduate students were recruited from the pool of level six students. The participants were randomly divided into the control and experimental groups. The control group consisted of 28 students; while the experimental group was comprised of 20 students. The 288 students’ writing productions were analyzed according to the fourteen measures of syntactic complex by using a paired t-test and an independent t test. For the first question, the results of this study show that there was no increase/decrease in syntactic complexity, either as a whole or partially across individual measures, for the control group for the three writing tasks (argumentation, classification, and reaction). This would suggest that traditional learning method does not reach to level to be significant to the participants in the control group. For the second research question, the t-tests showed that the syntactic complexity of the participants in the experimental group increased as a whole in both the classification and reaction essays. An increase in complexity was shown across the following measures for the classification essay: MLS, MLT, MLC, CN/T/VP/T, and CN/C. For the reaction essay, an increase in complexity was shown across MLS, MLT, MLC, and CN/T measures. Although the syntactic complexity of these participants did not increase/decrease as a whole for the argumentation essay, the CN/T measure did show some increase. Finally, the comparison between the results of the two groups revealed that, although the experimental group in this study showed more improvement in syntactic complexity than the control group, the degree of difference between the two groups was too small to draw any definite conclusion about the relative effectiveness of the two methods. This may be due to the comparatively short duration of the study: ten weeks. The findings of this research have significant implications for academic research and for Saudi EFL teachers at the university level.

ERANET JTC 2011: Submission and Activation of an International Academic Translational Project in Advanced Breast Cancer. Experience From the ET-FES Study

Background: Academic research is important to face unmet medical needs. The Oncological community encounters many hurdles in setting up multicenter investigator-driven trials mainly due to administrative complexity. The purpose of a network organization at a multinational level is to facilitate clinical trials through standardization, coordination, and education for drug development and regulatory approval.Methods: The application of an European grant foresees the creation of a consortium which aims at facilitating multi-center academic clinical trials.Results: The ERA-NET TRANSCAN Call 2011 on “Validation of biomarkers for personalized cancer medicine” was released on December 2011. This project included Italian, Spanish, French and German centers. The approval process included Consortium constitution, project submission, Clinical Trial Submission, and activation on a national level. The different timescales for submitting study documents in each Country and the misalignment of objections by each Competent Authority CA, generated several requests for changes to the study documents which meant amendments had to be made; as requested by the 2001/20/EC Directive, the alignment of core documents is mandatory. This procedure impacted significantly on study activation timelines. Time to first patient in was 14, 10, 28, and 31 months from the date of submission in Italy, France, Spain, and Germany, respectively. Accrual was stopped on 22nd January 2021 due to an 18F FES shortage as the primary reason but also for having exceeded the project deadlines with consequent exhaustion of the funds allocated for the project.Conclusions: Pharmaceutical companies might be reluctant to fund research projects aimed at treatment individualization if the approval for a wider indication has already been achieved. Academic trials therefore become fundamental for promoting trials which are not attractive to big pharma. It was very difficult and time consuming to activate an academic clinical trial, for this reason, a study may become “old” as new drugs entered into the market. National institutions should promote the development of clinical research infrastructures and network with competence in regulatory, ethical, and legal skills to speed up academic research.

Attitudes toward sex doll ownership: Scale development and preliminary validation

Sex dolls are beginning to become more mainstream, both in the public’s consciousness and in academic research. However, there is no current systematic examination of public attitudes toward sex dolls within the peer-reviewed literature, which represents a barrier to the efficient study on this topic. In this paper we report the development and initial validation of such a measure. Using an international public sample (N = 377) we found that public attitudes toward sex doll owners were underpinned by three factors: ‘Acceptability of Doll Ownership’, ‘Doll Owners as Immoral’, and ‘Doll Owners as Dysfunctional’. Scores on each of these factors, which make up the Sex Doll Ownership Attitudes Scale (SDOAS), were predicted by participant sex, religiosity, permissive sexual attitudes, right-wing authoritarianism, and the endorsement of moral intuitions that favor personal liberty. We present a full form 66-item version of the SDOAS, as well as an 18-item short form which both possess strong psychometric properties. We highlight potential future uses of the SDOAS as this emerging field of study continues to rapidly grow.

Scientific journals must be alert to potential manipulation in citations and referencing

Citation is an essential practice in scientific publishing. However, it is mandatory that citing the sources in a scientific work is performed in a proper manner. Manipulating citations in research articles is one form of academic research misconduct that violates publication ethics. Citation manipulation simply occurs for the purpose of increasing the number of citations of a researcher or a journal. Unfortunately, there has been a growing trend for this type of misconduct recently and this has not received much attention from the science community. The most effective solution to prevent the growth of such unethical practices is for reputable journals to impose stricter rules on reference evaluation criteria in order to emphasize on the appropriateness of the citations.

Intelligent Educational Evaluation of Research Performance between Digital Library and Open Government Data

This study evaluates institutional research performance in benchmark technological universities in Taiwan through intelligent research databases (SciVal) in digital libraries with Ministry of Education open data to explore the performance of research indicators and the research trend of topic clusters to ascertain accountability for decision makers. The research performance of eight benchmark technological universities in Taiwan is compared in this study. In addition, the trends in research topics in the top 10% of journals are explored. Descriptive statistics, correlation, ANOVA, and the Boston Consulting Group matrix were used in this study. Research personnel, publications, productivity, total citations, number of international collaborations, and academic research income in 2018 significantly positively correlated with each other. From 719 records of research topics, topic clusters and school types are the significant factors in research outputs. Biosensors, electrodes, and voltammetry are the leading topic clusters in the research trend. The topic cluster of decision-making, fuzzy sets, and models has the best growth rate in the SciVal results. This analysis provides useful insights to policymakers to improve institutional administration and research resource allocation.