Semantic Description of Explainable Machine Learning Workflows for Improving Trust

Explainable Machine Learning comprises methods and techniques that enable users to better understand the machine learning functioning and results. This work proposes an ontology that represents explainable machine learning experiments, allowing data scientists and developers to have a holistic view, a better understanding of the explainable machine learning process, and to build trust. We developed the ontology by reusing an existing domain-specific ontology (ML-SCHEMA) and grounding it in the Unified Foundational Ontology (UFO), aiming at achieving interoperability. The proposed ontology is structured in three modules: (1) the general module, (2) the specific module, and (3) the explanation module. The ontology was evaluated using a case study in the scenario of the COVID-19 pandemic using healthcare data from patients, which are sensitive data. In the case study, we trained a Support Vector Machine to predict mortality of patients infected with COVID-19 and applied existing explanation methods to generate explanations from the trained model. Based on the case study, we populated the ontology and queried it to ensure that it fulfills its intended purpose and to demonstrate its suitability.

Deep Reinforcement Learning Boosted Partial Domain Adaptation

Domain adaptation is critical for learning transferable features that effectively reduce the distribution difference among domains. In the era of big data, the availability of large-scale labeled datasets motivates partial domain adaptation (PDA) which deals with adaptation from large source domains to small target domains with less number of classes. In the PDA setting, it is crucial to transfer relevant source samples and eliminate irrelevant ones to mitigate negative transfer. In this paper, we propose a deep reinforcement learning based source data selector for PDA, which is capable of eliminating less relevant source samples automatically to boost existing adaptation methods. It determines to either keep or discard the source instances based on their feature representations so that more effective knowledge transfer across domains can be achieved via filtering out irrelevant samples. As a general module, the proposed DRL-based data selector can be integrated into any existing domain adaptation or partial domain adaptation models. Extensive experiments on several benchmark datasets demonstrate the superiority of the proposed DRL-based data selector which leads to state-of-the-art performance for various PDA tasks.

Development and Assessment of The Quality of Life Instruments for Chronic Diseases-Gout(QLICD-GO) (V2.0)

Background: To develop and assess the Quality of Life Instruments for Chronic Diseases-Gout (QLICD-GO [V2.0]).Methods: The instrument was developed using a programmatic decision-making method to combine the general module of the Quality of Life Instruments for Chronic Diseases and a new specific module. The instrument was assessed by measuring the quality of life of 116 patients with gout. Results: The QLICD-GO(V2.0) included 28 items from the general module of chronic diseases and 12 items in three facets from the specific module. In addition to the field of physiological function, the internal consistency reliability of other fields and dimensions of the instrument was >0.7 and the split-half reliability was >0.5. Three common factors were extracted from the specific module, with a cumulative variance contribution rate of 57.54%. The standardized response means of the specific module and the whole instrument were 0.94 and 1.20, respectively. Conclusions: The QLICD-GO(V2.0) has good reliability, validity, and responsiveness. The instrument comprehensively and objectively reflects the quality of life of patients with gout, and it can be used to assess treatment regimens developed by medical staff.

Development and Assessment of the Quality of Life Instruments for Chronic Diseases-Gout(QLICD-GO) (V2.0)

Objective To develop and assess the Quality of Life Instruments for Chronic Diseases-Gout (QLICD-GO [V2.0]).Methods The instrument was developed using a programmatic decision-making method to combine the general module of the Quality of Life Instruments for Chronic Diseases and a new specific module. The instrument was assessed by measuring the quality of life of 116 patients with gout.Results The QLICD-GO(V2.0) included 28 items from the general module of chronic diseases and 12 items in three facets from the specific module. In addition to the field of physiological function, the internal consistency reliability of other fields and dimensions of the instrument was > 0.7 and the split-half reliability was > 0.5. Three common factors were extracted from the specific module, with a cumulative variance contribution rate of 57.54%. The standardized response means of the specific module and the whole instrument were 0.94 and 1.20, respectively.Conclusions The QLICD-GO(V2.0) has good reliability, validity, and responsiveness. The instrument comprehensively and objectively reflects the quality of life of patients with gout, and it can be used to assess treatment regimens developed by medical staff.

Development and Evaluation of Quality of Life Instruments for Chronic Diseases–Chronic Pulmonary Heart Disease (QLICD-CPHD) (V2.0)

Objective: To develop a quality of life (QOL) instrument specific to chronic pulmonary heart disease (CPHD) and to evaluate its reliability, validity, and responsiveness. Methods: Under the Quality of Life Instruments for Chronic Diseases (QLICD) system, the QLICD-CPHD (V2.0) was developed and used to measure the QOL of 184 patients with CPHD. Data were analyzed by statistical description, linear correlation analysis, exploratory factor analysis, and paired t-test. Results: The QLICD-CPHD (V2.0) consisted of 44 items, including 28 in the general module and 16 in the specific module. The Cronbach’s α coefficients of four domains in the general module and the specific module were greater than 0.7, and the split-half reliability coefficients were between 0.6 and 0.8, indicating that the instrument has good reliability. The instrument showed good validity as the correlation coefficients among items of the same domain were significantly higher than those among items of different domains. Correlation coefficients r between the corresponding domains of SF-36 and QLICD-CPHD were between 0.26 and 0.60, which suggests that QLICD-CPHD has moderate criterion validity. Paired t-tests of scores of patients before and after treatment showed significant differences in the overall score, scores of general module domains except social role, and scores of the specific module. Standardized response means were greater than 0.2, suggesting that the instrument had moderate responsiveness and was sensitive to changes in the patients’ QOL. Conclusion: The QLICD-CPHD (V2.0) has good reliability, validity, and responsiveness. The instrument can be used to evaluate the QOL of CPHD patients.

Quality of life in children with infrequent congenital heart defects: cohort study with one-year of follow-up

Background The evidence regarding patient related outcomes in children with infrequent congenital heart defects (I-CHD) is very limited. We sought to measure quality of life (QoL) in children with I-CHD, and secondarily, to describe QoL changes after one-year of follow-up, self-reported by children and through their caregivers’ perspective. Methods We assembled a cohort of children diagnosed with an I-CHD in a cardiovascular referral center in Colombia, between August 2016 and September 2018. At baseline and at one-year follow-up, a clinical psychology assessment was performed to establish perception of QoL. The Pediatric Quality of Life Inventory (PedsQL) 4.0 scale was used in both general and cardiac modules for patients and for their caregivers. We used a Mann-Whitney U test to compare scores for general and cardiac modules between patients and caregivers, while a Wilcoxon test was used to compared patients’ and caregivers’ baseline and follow-up scores. Results are presented as median and interquartile range. Results To date, QoL evaluation at one-year follow-up has been achieved in 112/157 patients (71%). Self-reported scores in general and cardiac modules were higher than the QoL perceived through their caregivers, both at baseline and after one-year of follow-up. When compared, there was no statistically significant difference in general module scores at baseline between patients (median = 74.4, IQR = 64.1–80.4) and caregivers scores (median = 68.4, IQR = 59.6–83.7), p = 0.296. On the contrary, there was a statistical difference in baseline scores in the cardiac module between patients (median = 79.6, IQR = 69.7–87.4) and caregivers (median = 73.6, IQR = 62.6–84.3), p = 0.019. At one-year of follow-up, scores for the general module between patients (median = 72.8, IQR = 59.2–85.9) and caregivers (median = 69.9, IQR = 58.1–83.7) were not statistically different (p = 0.332). Finally, a significant difference was found for cardiac module scores between patient (median = 75.0, IQR = 67.1–87.1) and caregivers (median = 73.1, IQR = 59.5–83.8), p = 0.034. Conclusions QoL in children with I-CHD can be compromised. However, children have a better perception of their QoL when compared with their caregivers’ assessments. To provide high-quality care, besides a thorough clinical evaluation, QoL directly elicited by the child should be an essential aspect in the integral management of I-CHD.

The Effect of Impulsivity vs. Reflectivity on Reading Comprehension of Iranian EFL Learners

In recent years attention has been accorded to language learners’ affective factors and learning styles. Two of the significant learning styles are impulsivity and reflectivity which have not been studied as much as other styles such as introversion and extroversion. This study endeavored to find out whether or not impulsivity and reflectivity have any effect on reading comprehension of Iranian English as a Foreign Language (EFL) learners. Seventy two language learners were selected from 4 intact classes out of 112 learners. Nelson proficiency test was given to the participants as homogeneity test. Next, Barrat’s (1995) impulsiveness questionnaire was given to the participants. Based on the results of the questionnaire, the participants formed 3 different groups, i.e., a reflective group (n=25), impulsive group (n=25) and a control group (n=22). The control group consisted of less impulsive and less reflective learners based on Barrat’s scale. An IELTS reading test (general module) was administered to the participants. Based on the results of independent samples t-test, it was found that impulsivity and reflectivity do not have any effect on reading comprehension of Iranian EFL learners. Pedagogical implications are discussed.