A Data-Centric Approach to Improve Machine Learning Model’s Performance in Production

Machine learning teaches computers to think in a similar way to how humans do. An ML models work by exploring data and identifying patterns with minimal human intervention. A supervised ML model learns by mapping an input to an output based on labeled examples of input-output (X, y) pairs. Moreover, an unsupervised ML model works by discovering patterns and information that was previously undetected from unlabelled data. As an ML project is an extensively iterative process, there is always a need to change the ML code/model and datasets. However, when an ML model achieves 70-75% of accuracy, then the code or algorithm most probably works fine. Nevertheless, in many cases, e.g., medical or spam detection models, 75% accuracy is too low to deploy in production. A medical model used in susceptible tasks such as detecting certain diseases must have an accuracy label of 98-99%. Furthermore, that is a big challenge to achieve. In that scenario, we may have a good working model, so a model-centric approach may not help much achieve the desired accuracy threshold. However, improving the dataset will improve the overall performance of the model. Improving the dataset does not always require bringing more and more data into the dataset. Improving the quality of the data by establishing a reasonable baseline level of performance, labeler consistency, error analysis, and performance auditing will thoroughly improve the model's accuracy. This review paper focuses on the data-centric approach to improve the performance of a production machine learning model.

How does working memory store more information at larger set sizes? A composite code model.

A central feature of working memory is its limited capacity in terms of the amount of information that can be simultaneously maintained. Despite this, many studies observe an increase in the total amount when more items are maintained (set size), as measured by Shannon information. We propose the composite code model which maintains this fixed capacity assumption but demonstrates increasing observed information across set sizes. This relies on the hierarchical organisation of the visual system, in which higher-order information is abstracted about simple study displays. Using Bayesian inference, target responses can be inferred from knowledge about non-targets. We tested this model against our own data from a delayed reproduction task and those of published open data sets. We found initial support for the model, with its predictions matching those of the observed effects.

How does working memory store more information at larger set sizes? A composite code model.

A central feature of working memory is its limited capacity in terms of the amount of information that can be simultaneously maintained. Despite this, many studies observe an increase in the total amount when more items are maintained (set size), as measured by Shannon information. We propose the composite code model which maintains this fixed capacity assumption but demonstrates increasing observed information across set sizes. This relies on the hierarchical organisation of the visual system, in which higher-order information is abstracted about simple study displays. Using Bayesian inference, target responses can be inferred from knowledge about non-targets. We tested this model against our own data from a delayed reproduction task and those of published open data sets. We found initial support for the model, with its predictions matching those of the observed effects.

Multiverse analyses in fear conditioning research

There is heterogeneity in and a lack of consensus on the preferred statistical analyses foranalyzing fear conditioning effects in light of a multitude of potentially equally justifiablestatistical approaches. Here, we introduce the concept of multiverse analysis for fearconditioning research. We also present a model multiverse approach specifically tailored tofear conditioning research and introduce the novel and easy to use R package ‘multifear’ thatallows to run all the models though a single line of code. Model specifications and datareduction approaches employed in the ‘multifear’ package were identified through arepresentative systematic literature search. The heterogeneity of statistical models identifiedincluded Bayesian ANOVA and t-tests as well as frequentist ANOVA, t-test as well as mixedmodels with a variety of data reduction approaches (i.e., number of trials, trial blocks,averages) as input. We illustrate the power of a multiverse analysis for fear conditioning databased on two pre-existing data sets with partial (data set 1) and 100% reinforcement rate(data set 2) by using CS discrimination in skin conductance responses (SCRs) during fearacquisition and extinction training as case examples. Both the effect size and the direction ofeffect was impacted by choice of the model and data reduction techniques. We anticipatethat an increase in multiverse-type of studies in the field of fear conditioning research andtheir extension to other outcome measures as well as data and design multiverse analyseswill aid the development of formal theories through the accumulation of empirical evidence.This may contribute to facilitated and more successful clinical translation.

Exploring the Motivation of Pupils towards the Implementation of QR Codes in Pronunciation Learning

The implementation of Quick Response (QR) codes in education has become increasingly popular. Previous studies proved that it can increase the motivation of pupils in learning. Hence, this research investigates the difference of pupils’ motivation levels in learning pronunciation after the implementation of QR codes. We examine the motivation of pupils in terms of four aspects: interest, competence, perceived choice and sense of belonging. Accordingly, we propose a QR code model, which is linked to Google Forms that contain audio recordings for pronunciation practice. With the proposed QR code activity, the pupils can record their versions of audio recordings of pronunciation and submit to their teacher for feedback. The participants are 90 year 4 pupils from a sub-urban Chinese primary school in Johor. We employ a pre-experimental research design in this research. We collect our data by using 2 research instruments: survey questionnaires and observation checklist. We administer the survey questionnaires before and after the implementation of QR codes to find out the changes in the motivation of pupils. We also utilise an observation checklist to examine the attitude of the participants during the implementation of QR codes in pronunciation learning. The findings of this research reveal a significant change in the pupils’ motivation towards the implementation of QR codes in pronunciation learning. Specifically, we find an increment of motivation in learning pronunciation as the pupils show interests in learning. Received: 6 October 2020 / Accepted: 12 December 2020 / Published: 17 January 2021

Towards a Generation of Class Diagram From User Stories in Agile Methods

Model-driven architecture (MDA) is an alternative approach of software engineering that allows an automatic transformation from business process model to code model. In MDA there are two transformation kinds: transformation from computing independent model (CIM) to platform independent model (PIM) and transformation from PIM to platform specific model (PSM). In this chapter, the authors based on CIM to PIM transformation. This transformation is done by developing a platform that generates class diagram, presented in XMI file, from specifications that are presented in user stories, which are written in natural language (English). They used a natural language processing (NLP) tool named “Stanford Core NLP” for extracting of the object-oriented design elements. The approach was validated by focusing on two case studies: firstly, comparing the results with the results other researchers; and secondly, comparing the results with the results obtained manually. The benefits of the approach are aligned with agile methods goals.

Defects in the 3-dimensional toric code model form a braided fusion 2-category

It was well known that there are e-particles and m-strings in the 3-dimensional (spatial dimension) toric code model, which realizes the 3-dimensional ℤ2 topological order. Recent mathematical result, however, shows that there are additional string-like topological defects in the 3-dimensional ℤ2 topological order. In this work, we construct all topological defects of codimension 2 and higher, and show that they form a braided fusion 2-category satisfying a braiding non-degeneracy condition.

Atomicity Violation in Multithreaded Applications and Its Detection in Static Code Analysis Process

This paper is a contribution to the field of research dealing with the parallel computing, which is used in multithreaded applications. The paper discusses the characteristics of atomicity violation in multithreaded applications and develops a new definition of atomicity violation based on previously defined relationships between operations, that can be used to atomicity violation detection. A method of detection of conflicts causing atomicity violation was also developed using the source code model of multithreaded applications that predicts errors in the software.