We focus on multi-dimensional classification (MDC) problems with conditional dependence, which we call multiple output dependence (MOD) problems. MDC is the task of predicting a vector of categorical outputs for each input. Conditional dependence in MDC means that the choice for one output value affects the choice for others, so it is not desirable to predict outputs independently. We show that conditional dependence in MDC implies that a single input can map to multiple correct output vectors. This means it is desirable to find multiple correct output vectors per input. Current solutions for MOD problems are not sufficient because they predict only one of the correct output vectors per input, ignoring all others.We modify four existing MDC solutions, including chain classifiers, to predict multiple output vectors. We further create a novel ensemble technique named weighted output vector ensemble (WOVE) which combines these multiple predictions from multiple chain classifiers in a way that preserves the integrity of output vectors and thus preserves conditional dependence among outputs. We verify the effectiveness of WOVE by comparing it against 7 other solutions on a variety of data sets and find that it shows significant gains over existing methods.

We compare the transition barrier that accompanies a first-order phase transition in the canonical and microcanonical ensemble. This is directly encoded in the probability distributions of standard Metropolis Monte Carlo simulations and a proper microcanonical sampling technique. For the example of droplet formation, we find that in both ensembles the transition barrier scales as expected but that the barrier is much smaller in the microcanonical ensemble. In addition its growth with system size is weaker which will enhance this difference for larger systems. We provide an intuitive physical explanation for this observation

No description available.

[pt] Atualmente algoritmos de aprendizado de máquina (ML) estão incrivelmente presentes no nosso cotidiano, desde sistemas de recomendação de filmes e músicas até áreas de alto risco como saúde, justiça criminal, finanças e assim por diante, auxiliando na tomada de decisões. Mas a complexidade de criação desses algoritmos de ML também está aumentando, enquanto sua interpretabilidade está diminuindo. Muitos algoritmos e suas decisões não podem ser facilmente explicados por desenvolvedores ou usuários, e os algoritmos também não são autoexplicáveis. Com isso, erros e vieses podem acabar ficando ocultos, o que pode impactar profundamente a vida das pessoas. Devido a isso, iniciativas relacionadas a transparência, explicabilidade e interpretabilidade estão se tornando cada vez mais relevantes, como podemos ver no novo regulamento sobre proteção e tratamento de dados pessoais (GDPR, do inglês General Data Protection Regulation), aprovado em 2016 para a União Europeia, e também na Lei Geral de Proteção de Dados (LGPD) aprovada em 2020 no Brasil. Além de leis e regulamentações tratando sobre o tema, diversos autores consideram necessário o uso de algoritmos inerentemente interpretáveis; outros mostram alternativas para se explicar algoritmos caixa-preta usando explicações locais, tomando a vizinhança de um determinado ponto e então analisando a fronteira de decisão dessa região; enquanto ainda outros estudam o uso de explicações contrafactuais. Seguindo essa linha dos contrafactuais, nos propomos a desenvolver uma interface com usuário para o sistema Optimal Counterfactual Explanations in Tree Ensembles (OCEAN), denominada OceanUI, através do qual o usuário gera explicações contrafactuais plausíveis usando Programação Inteira Mista e Isolation Forest. O propósito desta interface é facilitar a geração de contrafactuais e permitir ao usuário obter um contrafactual personalizado e mais aplicável individualmente, por meio da utilização de restrições e gráficos interativos. / [en] Machine learning algorithms (ML) are becoming incredibly present in our daily lives, from movie and song recommendation systems to high-risk areas like health care, criminal justice, finance, and so on, supporting decision making. But the complexity of those algorithms is increasing while their interpretability is decreasing. Many algorithms and their decisions cannot be easily explained by either developers or users, and the algorithms are also not self-explanatory. As a result, mistakes and biases can end up being hidden, which can profoundly impact people s lives. So, initiatives concerning transparency, explainability, and interpretability are becoming increasingly more relevant, as we can see in the General Data Protection Regulation (GDPR), approved in 2016 for the European Union, and in the General Data Protection Law (LGPD) approved in 2020 in Brazil. In addition to laws and regulations, several authors consider necessary the use of inherently interpretable algorithms; others show alternatives to explain black-box algorithms using local explanations, taking the neighborhood of a given point and then analyzing the decision boundary in that region; while yet others study the use of counterfactual explanations. Following the path of counterfactuals, we propose to develop a user interface for the system Optimal Counterfactual Explanations in Tree Ensembles (OCEAN), which we call OceanUI, through which the user generates plausible counterfactual explanations using Mixed Integer Programming and Isolation Forest. The purpose of this user interface is to facilitate the counterfactual generation and to allow the user to obtain a personal and more individually applicable counterfactual, by means ofrestrictions and interactive graphics.

Churn prediction methods are widely used in Customer Relationship Management and have proven to be valuable for retaining customers. To obtain a high predictive performance, recent studies rely on increasingly complex machine learning methods, such as ensemble or hybrid models. However, the more complex a model is, the more difficult it becomes to understand how decisions are actually made. Previous studies on machine learning interpretability have used a global perspective for understanding black-box models. This study explores the use of local explanation models for explaining the individual predictions of a Random Forest ensemble model. The churn prediction was studied on the users of Tink – a finance app. This thesis aims to take local explanations one step further by making comparisons between churn indicators of different user groups. Three sets of groups were created based on differences in three user features. The importance scores of all globally found churn indicators were then computed for each group with the help of local explanation models. The results showed that the groups did not have any significant differences regarding the globally most important churn indicators. Instead, differences were found for globally less important churn indicators, concerning the type of information that users stored in the app. In addition to comparing churn indicators between user groups, the result of this study was a well-performing Random Forest ensemble model with the ability of explaining the reason behind churn predictions for individual users. The model proved to be significantly better than a number of simpler models, with an average AUC of 0.93. / Metoder för att prediktera utträde är vanliga inom Customer Relationship Management och har visat sig vara värdefulla när det kommer till att behålla kunder. För att kunna prediktera utträde med så hög säkerhet som möjligt har den senasteforskningen fokuserat på alltmer komplexa maskininlärningsmodeller, såsom ensembler och hybridmodeller. En konsekvens av att ha alltmer komplexa modellerär dock att det blir svårare och svårare att förstå hur en viss modell har kommitfram till ett visst beslut. Tidigare studier inom maskininlärningsinterpretering har haft ett globalt perspektiv för att förklara svårförståeliga modeller. Denna studieutforskar lokala förklaringsmodeller för att förklara individuella beslut av en ensemblemodell känd som 'Random Forest'. Prediktionen av utträde studeras påanvändarna av Tink – en finansapp. Syftet med denna studie är att ta lokala förklaringsmodeller ett steg längre genomatt göra jämförelser av indikatorer för utträde mellan olika användargrupper. Totalt undersöktes tre par av grupper som påvisade skillnader i tre olika variabler. Sedan användes lokala förklaringsmodeller till att beräkna hur viktiga alla globaltfunna indikatorer för utträde var för respektive grupp. Resultaten visade att detinte fanns några signifikanta skillnader mellan grupperna gällande huvudindikatorerna för utträde. Istället visade resultaten skillnader i mindre viktiga indikatorer som hade att göra med den typ av information som lagras av användarna i appen. Förutom att undersöka skillnader i indikatorer för utträde resulterade dennastudie i en välfungerande modell för att prediktera utträde med förmågan attförklara individuella beslut. Random Forest-modellen visade sig vara signifikantbättre än ett antal enklare modeller, med ett AUC-värde på 0.93.

The density of single grain states in static granular packings is derived from first principles for an idealized yet fundamental case. This produces the distribution of contact forces P_f(f) in the packing. Because there has been some controversy in the published literature over the exact form of the distribution, this dissertation begins by reviewing the existing empirical observations to resolve those controversies. A method is then developed to analyze Edwards' granular contact force probability functional from first principles. The derivation assumes Edwards' flat measure -- a density of states (DOS) that is uniform within the metastable regions of phase space. A further assumption, supported by physical arguments and empirical evidence, is that contact force correlations arising through the closure of loops of grains may be neglected. Then, maximizing a state-counting entropy results in a transport equation that can be solved numerically. For the present it has been solved using the "Mean Structure Approximation," projecting the DOS across all angular coordinates to more clearly identify its predominant features in the remaining stress coordinates. These features are: (1) the Grain Factor related to grain stability and strong correlation between the contact forces on the same grain, and (2) the Structure Factor related to Newton's third law and strong correlation between neighboring grains. Numerical simulations were then performed for idealized granular packings to permit a direct comparison with the theory, and the data including P_f(f) were found to be in excellent agreement. Where the simulations and theory disagree, it is primarily due to the coordination number Z because the theory assumes Z to be a constant whereas in disordered packings it is not. The form of the empirical DOS is discovered to have an elegant, underlying pattern related to Z. This pattern consists entirely of the functional forms correctly predicted by the theory, but with only slight parameter changes as a function of Z. This produces significant physical insight and suggests how the theory may be generalized in the future.

This text provides an overview of problems in the field of data assimilation. We explore the possibility of recreating unknown data by continuously inserting known data into certain dynamical systems, under certain regularity assumptions. Additionally, we discuss an alternative statistical approach to data assimilation and investigate the utilization of the Ensemble Kalman Filter for assimilating data into dynamical models. A key challenge in numerical weather prediction is incorporating convective precipitation into an idealized setting for numerical computations. To answer this question we examine the modified rotating shallow water equations, a nonlinear coupled system of partial differential equations and further assess if this primitive model accurately mimics phenomena observed in operational numerical weather prediction models. Numerical experiments conducted using a Deterministic Ensemble Kalman Filter algorithm support its applicability for convective-scale data assimilation. Furthermore, we analyze the frequency spectrum of numerical forecasts using the Wavelet transform. Our frequency analysis suggests that, under certain experimental settings, there are similarities in the initialization of operational models, which can aid in understanding the problem of intialization of numerical weather prediction models.

One-class classification is a specialized form of classification from the field of machine learning. Traditional classification attempts to assign unknowns to known classes, but cannot handle novel unknowns that do not belong to any of the known classes. One-class classification seeks to identify these outliers, while still correctly assigning unknowns to classes appropriately. One-class classification is applied here to the field of nuclear forensics, which is the study and analysis of nuclear material for the purpose of nuclear incident investigations. Nuclear forensics data poses an interesting challenge because false positive identification can prove costly and data is often small, high-dimensional, and sparse, which is problematic for most machine learning approaches. A web application is built using the R programming language and the shiny framework that incorporates N-SLOPE: a machine learning ensemble. N-SLOPE combines five existing one-class classifiers with a novel one-class classifier introduced here and uses ensemble learning techniques to combine output. N-SLOPE is validated on three distinct data sets: Iris, Obsidian, and Galaxy Serpent 3, which is an enhanced version of a recent international nuclear forensics exercise. N-SLOPE achieves high classification accuracy on each data set of 100%, 83.33%, and 83.33%, respectively, while minimizing false positive detection rate to 0% across the board and correctly detecting every single novel unknown from each data set. N-SLOPE is shown to be a useful and powerful tool to aid in nuclear forensic investigations.

Physical activity can have immediate and long-term benefits on health and reduce the risk for chronic diseases. Valid measures of physical activity are needed in order to improve our understanding of the exact relationship between physical activity and health. Activity monitors have become a standard for measuring physical activity; accelerometers in particular are widely used in research and consumer products because they are objective, inexpensive, and practical. Previous studies have experimented with different monitor placements and classification methods. However, the majority of these methods were developed using data collected in controlled, laboratory-based settings, which is not reliably representative of real life data. Therefore, more work is required to validate these methods in free-living settings. For our work, 25 participants were directly observed by trained observers for two two-hour activity sessions over a seven day timespan. During the sessions, the participants wore accelerometers on the wrist, thigh, and chest. In this thesis, we tested a battery of machine learning techniques, including a hierarchical classification schema and a confusion matrix boosting method to predict activity type, activity intensity, and sedentary time in one-second intervals. To do this, we created a dataset containing almost 100 hours worth of observations from three sets of accelerometer data from an ActiGraph wrist monitor, a BioStampRC thigh monitor, and a BioStampRC chest monitor. Random forest and k-nearest neighbors are shown to consistently perform the best out of our traditional machine learning techniques. In addition, we reduce the severity of error from our traditional random forest classifiers on some monitors using a hierarchical classification approach, and combat the imbalanced nature of our dataset using a multi-class (confusion matrix) boosting method. Out of the three monitors, our models most accurately predict activity using either or both of the BioStamp accelerometers (with the exception of the chest BioStamp predicting sedentary time). Our results show that we outperform previous methods while still predicting behavior at a more granular level.

In the context of global warming, to reduce energy consumption, an unavoidable policy is to renovate badly-isolated buildings. However, most studies concerning efficiency of renovation work do not rely on energy data from smart meters but rather on estimates. To develop a precise tool to assess the quality of renovation work, several ensemble models were tested and compared with existing ones. Each model learns the consumption habits before the date of the works and then predicts what the energy load curve would have been if the works had not been realized. The prediction is finally compared to the actual energy load to infer the savings over the same dataset. The models were compared using precision and time complexity metrics. The best ensemble model’s precision scores are equivalent to the state-of-the-art. Moreover, the developed model is 32 times quicker to fit and predict. / I samband med den globala uppvärmningen är det oundvikligt att renovera dåligt isolerade byggnader för att minska energiförbrukningen. De flesta studier om renoveringsarbetenas effektivitet bygger dock inte på energidata från smarta mätare utan snarare på uppskattningar. För att utveckla ett exakt verktyg för att bedöma kvaliteten på renoveringsarbeten har flera ensemblemodeller testats och jämförts med befintliga modeller. Varje modell lär sig förbrukningsvanorna före arbetena och förutspår sedan hur energibelastningskurvan skulle ha sett ut om arbetena inte hade genomförts. Prognosen jämförs slutligen med den faktiska energilasten för att härleda besparingarna för samma dataset. Modellerna jämfördes med hjälp av precision och tidskomplexitet. Den bästa ensemblemodellens precisionspoäng är likvärdig med den bästa modellen. Dessutom är den utvecklade modellen 32 gånger snabbare att anpassa och förutsäga