High Degree Of Interpretability Research Articles

Energy consumption forecasting in PCM-integration buildings considering building and environmental parameters for future climate scenarios

Despite numerous machine learning methods being employed to predict the energy consumption of PCM-integrated buildings, key research gaps remain. Most studies focus solely on building parameters while omitting important environmental parameters, including precipitation and air pressure. No study evaluated and proposed prediction models for PCM-integrated buildings considering future climate scenarios. Also, as per the authors′ knowledge, no researcher has assessed the impact of variations in the hyperparameter, especially for decision tree-based prediction models to develop a reliable prediction model with less complexity and a high degree of interpretability between independent and dependent variables. This research addresses these gaps by evaluating fine, medium, and coarse decision trees for predicting energy consumption in PCM-integrated buildings under future climate scenarios by considering extensive building and environmental parameters simultaneously. A database for energy consumption was created through energy simulations for 11 cities in hot semi-arid climates. The Fine Decision Tree (FDT3) emerged as the most accurate prediction model, with R2 values over 94 % in training and testing phases based on model evaluation and validation processes. Parametric analysis further revealed that both environmental and building parameters are crucial in accurately predicting the energy consumption of PCM-integrated buildings using FDT3.

GradTree: Learning Axis-Aligned Decision Trees with Gradient Descent

Decision Trees (DTs) are commonly used for many machine learning tasks due to their high degree of interpretability. However, learning a DT from data is a difficult optimization problem, as it is non-convex and non-differentiable. Therefore, common approaches learn DTs using a greedy growth algorithm that minimizes the impurity locally at each internal node. Unfortunately, this greedy procedure can lead to inaccurate trees. In this paper, we present a novel approach for learning hard, axis-aligned DTs with gradient descent. The proposed method uses backpropagation with a straight-through operator on a dense DT representation, to jointly optimize all tree parameters. Our approach outperforms existing methods on binary classification benchmarks and achieves competitive results for multi-class tasks. The implementation is available under: https://github.com/s-marton/GradTree

XAS: Automatic yet eXplainable Age and Sex determination by combining imprecise per-tooth predictions

Chronological age and biological sex estimation are two key tasks in a variety of procedures, including human identification and migration control. Issues such as these have led to the development of both semiautomatic and automatic prediction models, but the former are expensive in terms of time and human resources, while the latter lack the interpretability required to be applicable in real-life scenarios. This paper therefore proposes a new, fully automatic methodology for the estimation of age and sex. This first applies a tooth detection by means of a modified CNN with the objective of extracting the oriented bounding boxes of each tooth. Then, it feeds the image features inside the tooth boxes into a second CNN module designed to produce per-tooth age and sex probability distributions. The method then adopts an uncertainty-aware policy to aggregate these estimated distributions. Our approach yielded a lower mean absolute error than any other previously described, at 0.97 years. The accuracy of the sex classification was 91.82%, confirming the suitability of the teeth for this purpose. The proposed model also allows analyses of age and sex estimations on every tooth, enabling experts to identify the most relevant for each task or population cohort or to detect potential developmental problems. In conclusion, the performance of the method in both age and sex predictions is excellent and has a high degree of interpretability, making it suitable for use in a wide range of application scenarios.

DAISY: An Implementation of Five Core Principles for Transparent and Accountable Conversational AI

We present a detailed implementation of five core principles for transparent and acccountable conversational AI, namely interpretability, inherent capability to explain, independent data, interactive learning, and inquisitiveness. This implementation is a dialogue manager called DAISY that serves as the core part of a conversational agent. We show how DAISY-based agents are trained with human-machine interaction, a process that also involves suggestions for generalization from the agent itself. Moreover, these agents are capable to provide a concise and clear explanation of the actions required to reach a conclusion. Deep neural networks (DNNs) are currently the de facto standard in conversational AI. We therefore formulate a comparison between DAISY-based agents and two methods that use DNNs, on two popular data sets involving multi-domain task-oriented dialogue. Specifically, we provide quantitative results related to entity retrieval and qualitative results in terms of the type of errors that may occur. The results show that DAISY-based agents achieve superior precision at the price of lower recall, an outcome that might be preferable in task-oriented settings. Ultimately, and especially in view of their high degree of interpretability, DAISY-based agents are a fundamentally different alternative to the currently popular DNN-based methods.

Personalized seizure signature: An interpretable approach to false alarm reduction for long-term epileptic seizure detection.

Long-term automatic detection of focal seizures remains one of the major challenges in epilepsy due to the unacceptably high number of false alarms from state-of-the-art methods. Our aim was to investigate to what extent a new patient-specific approach based on similarly occurring morphological electroencephalographic (EEG) signal patterns could be used to distinguish seizures from nonseizure events, as well as to estimate its maximum performance. We evaluated our approach on >5500h of long-term EEG recordings using two public datasets: the PhysioNet.org Children's Hospital Boston-Massachusetts Institute of Technology (CHB-MIT) Scalp EEG database and the EPILEPSIAE European epilepsy database. We visually identified a set of similarly occurring morphological patterns (seizure signature) seen simultaneously over two different EEG channels, and within two randomly selected seizures from each individual. The same seizure signature was then searched for in the entire recording from the same patient using dynamic time warping (DTW) as a similarity metric, with a threshold set to reflect the maximum sensitivity our algorithm could achieve without false alarm. At a DTW threshold providing no false alarm during the entire recordings, the mean seizure detection sensitivity across patients was 84%, including 96% for the CHB-MIT database and 74% for the European epilepsy database. A 100% sensitivity was reached in 50% of patients, including 79% from the CHB-MIT database and 27% from the European epilepsy database. The median latency from seizure onset to its detection was 17±10s, with 84% of seizures being detected within 40s. Personalized EEG signature combined with DTW appears to be a promising method to detect ictal events from a limited number of EEG channels with high sensitivity despite low rate of false alarms, high degree of interpretability, and low computational complexity, compatible with its future use in wearable devices.

The Effect of Online Learning on Learning Interest During The Covid-19 Pandemic

The purpose of this study was to ascertain students' interest in learning during their time spent online. The method used in this study is ex post facto quantitative research. The population of this study was approximately 307 students, which corresponds to the total number of PGSD fourth-semester students. This study's sample consisted of 40 individuals. A researcher gathered data via a google form questionnaire. The descriptive and inferential data analysis techniques used in this study are descriptive and inferential. The study's findings indicate a 0.562-degree relationship between online study and student interest in learning. This is the degree to which a moderate correlation exists with a positive relationship. With a 0.05 significance level, the result of rcount = 0.562 from N = 40 and a 1% error level, it can be stated that rcount rtable. As a result of this survey, it has been determined that there is a correlation between online study and student interest-based learning. With a high degree of interpretation, the online study does not affect student motivation to learn.

Demonstration and Mitigation of Spatial Sampling Bias for Machine-Learning Predictions

Summary Machine learning provides powerful methods for inferential and predictive modeling of complicated multivariate relationships to support decision-making for spatial problems such as optimization of unconventional reservoir development. Current machine-learning methods have been widely used in exhaustive spatial data sets like satellite images. However, geological subsurface characterization is significantly different because it is conditioned by sparse, nonrepresentative sampling. These sparse spatial data sets are generally not sampled in a representative manner; therefore, they are biased. The critical questions are: first, does spatial bias in training data result in a bias for machine-learning-based predictive models; and if there is a bias, how can we mitigate the bias in these spatial machine-learning-based predictions? The presence and mitigation of prediction with spatial sampling bias is demonstrated with tree-based machine learning due to its high degree of interpretability. In expectation, training data bias imposes bias in machine-learning predictions over a wide variety of spatial data configurations and degrees of bias, even when the model is applied to make predictions with unbiased testing and real-world data. We reduce the bias in prediction with a novel spatial weighted tree method over a variety of spatial data configurations and degrees of spatial sampling bias. The proposed method is able to improve the accuracy for reservoir evaluation. We recommend modeling checking and bias mitigation for all machine-learning prediction models with sparse, spatial data sets, because bias in, bias out.

Learning the grain boundary manifold: tools for visualizing and fitting grain boundary properties

With the proliferation of grain boundary data in materials science from both experiments and simulations, tools are needed to explore the five dimensional space of grain boundaries and visualize and fit structure property relationships along arbitrary paths through this space. In this work, we leverage a recently developed geodesic metric for grain boundaries to visualize the global geometry of grain boundary datasets and fit grain boundary energy to macroscopic grain boundary geometry. It is found that the 5D connectivity of the 388 grain boundary Olmsted dataset can be visualized via dimensionality reduction in 3D with a high degree of interpretability. Furthermore, after selectively adding new grain boundaries to the dataset, these visualizations suggest new global features of grain boundary space, including the existence of a grain boundary fundamental zone with well defined subsets of high symmetry boundaries along faces. Geodesic sampling is shown to be an effective tool to extend grain boundary datasets to new regions of the 5D space. Finally, a simple grain boundary energy kernel regression model with only one fitting parameter is demonstrated to predict grain boundary energy in the Olmsted dataset to within 10% RMSE.

Shape-based outlier detection in multivariate functional data

Multivariate functional data refer to a population of multivariate functions generated by a system involving dynamic parameters depending on continuous variables (e.g., multivariate time series). Outlier detection in such a context is a challenging problem because both the individual behavior of the parameters and the dynamic correlation between them are important. To address this problem, recent work has focused on multivariate functional depth to identify the outliers in a given dataset. However, most previous approaches fail when the outlyingness manifests itself in curve shape rather than curve magnitude. In this paper, we propose identifying outliers in multivariate functional data by a method whereby different outlying features are captured based on mapping functions from differential geometry. In this regard, we extract shape features reflecting the outlyingness of a curve with a high degree of interpretability. We conduct an experimental study on real and synthetic datasets and compare the proposed method with functional-depth-based methods. The results demonstrate that the proposed method, combined with state-of-the-art outlier detection algorithms, can outperform the functional-depth-based methods. Moreover, in contrast with the baseline methods, it is efficient regardless of the proportion of outliers.

Hybrid decision trees for data streams based on Incremental Flexible Naive Bayes prediction at leaf nodes

Mining data over streams in one pass and using constant memory is a challenging task. Decision trees are one of the most popular classifiers for both batch and incremental learning due to their high degree of interpretability, ease of construction and good accuracy. The most popular decision tree for stream classification is Hoeffding Tree based on Hoeffding bound. Literature shows a few variants of decision trees based on different bounds. The default class prediction method adopted in decision tree is “majority class” approach. Later, the accuracy of prediction was scaled up by a hybrid decision tree where Naive Bayes classifier was used for prediction. Kernel Density Estimation (KDE) is employed in Flexible Naive Bayes for classification. However, it is suitable for modeling static data set. This paper proposes an Incremental Flexible Naive Bayes (IFNB) based hybrid decision tree paradigm that uses KDE to model continuous attributes at leaf nodes of the tree for improving the class prediction accuracy. Experimental results on both synthetic and real dataset show that the proposed IFNB based leaf classifiers achieves improvement over the class prediction methods adopted in existing decision trees for data streams.

Open Source Bayesian Models. 3. Composite Models for Prediction of Binned Responses.

Bayesian models constructed from structure-derived fingerprints have been a popular and useful method for drug discovery research when applied to bioactivity measurements that can be effectively classified as active or inactive. The results can be used to rank candidate structures according to their probability of activity, and this ranking benefits from the high degree of interpretability when structure-based fingerprints are used, making the results chemically intuitive. Besides selecting an activity threshold, building a Bayesian model is fast and requires few or no parameters or user intervention. The method also does not suffer from such acute overtraining problems as quantitative structure–activity relationships or quantitative structure–property relationships (QSAR/QSPR). This makes it an approach highly suitable for automated workflows that are independent of user expertise or prior knowledge of the training data. We now describe a new method for creating a composite group of Bayesian models to extend the method to work with multiple states, rather than just binary. Incoming activities are divided into bins, each covering a mutually exclusive range of activities. For each of these bins, a Bayesian model is created to model whether or not the compound belongs in the bin. Analyzing putative molecules using the composite model involves making a prediction for each bin and examining the relative likelihood for each assignment, for example, highest value wins. The method has been evaluated on a collection of hundreds of data sets extracted from ChEMBL v20 and validated data sets for ADME/Tox and bioactivity.

Understanding large text corpora via sparse machine learning

AbstractSparse machine learning has recently emerged as powerful tool to obtain models of high‐dimensional data with high degree of interpretability, at low computational cost. The approach has been successfully used in many areas, such as signal and image processing. This article posits that these methods can be extremely useful in the analysis of large collections of text documents, without requiring user expertise in machine learning. Our approach relies on three main ingredients: (i) multidocument text summarization; (ii) comparative summarization of two corpora, both using sparse regression or classification; (iii) sparse principal components and sparse graphical models for unsupervised analysis and visualization of large text corpora. We validate our methods using a corpus of Aviation Safety Reporting System (ASRS) reports and demonstrate that the methods can reveal causal and contributing factors in runway incursions. Furthermore, we show that the methods automatically discover four main tasks that pilots perform during flight, which can aid in further understanding the causal and contributing factors to runway incursions and other drivers for aviation safety incidents. We also provide a comparative study involving other commonly used datasets, and report on the competitiveness of sparse machine learning compared to state‐of‐the‐art methods such as latent Dirichlet allocation (LDA). © 2013 Wiley Periodicals, Inc. Statistical Analysis and Data Mining 6: 221–242, 2013