Although few-shot learning and one-class classification (OCC), i.e. learning a binary classifier with data from only one class, have been separately well studied, their intersection remains rather unexplored. Our work addresses the few-shot OCC problem and presents a method to modify the episodic data sampling strategy of the model-agnostic meta-learning (MAML) algorithm to learn a model initialization particularly suited for learning few-shot OCC tasks. This is done by explicitly optimizing for an initialization which only requires few gradient steps with one-class minibatches to yield a performance increase on class-balanced test data. We provide a theoretical analysis that explains why our approach works in the few-shot OCC scenario, while other meta-learning algorithms, including MAML, fail. Our experiments on eight datasets from the image and time-series domains show that our method leads to higher results than classical OCC and few-shot classification approaches, and demonstrate the ability to learn unseen tasks from only few normal class samples. Moreover, we successfully train anomaly detectors for a real-world application on sensor readings recorded during industrial manufacturing of workpieces with a CNC milling machine using few normal examples. Finally, we empirically demonstrate that the proposed data sampling technique increases the performance of more recent meta-learning algorithms in few-shot OCC.

Modern requirements for machine learning (ML) models include both high predictive performance and model interpretability. A growing number of techniques provide model interpretations, but can lead to wrong conclusions if applied incorrectly. We illustrate pitfalls of ML model interpretation such as bad model generalization, dependent features, feature interactions or unjustified causal interpretations. Our paper addresses ML practitioners by raising awareness of pitfalls and pointing out solutions for correct model interpretation, as well as ML researchers by discussing open issues for further research.

Automated Machine Learning, which supports practitioners and researchers with the tedious task of manually designing machine learning pipelines, has recently achieved substantial success. In this paper we introduce new Automated Machine Learning (AutoML) techniques motivated by our winning submission to the second ChaLearn AutoML challenge, PoSH Auto-sklearn. For this, we extend Auto-sklearn with a new, simpler meta-learning technique, improve its way of handling iterative algorithms and enhance it with a successful bandit strategy for budget allocation. Furthermore, we go one step further and study the design space of AutoML itself and propose a solution towards truly hand-free AutoML. Together, these changes give rise to the next generation of our AutoML system, Auto-sklearn (2.0). We verify the improvement by these additions in a large experimental study on 39 AutoML benchmark datasets and conclude the paper by comparing to Auto-sklearn (1.0), reducing the regret by up to a factor of five.

A default assumption in many machine learning scenarios is that the training and test samples are drawn from the same probability distribution. However, such an assumption is often violated in the real world due to non-stationarity of the environment or bias in sample selection. In this work, we consider a prevalent setting called covariate shift, where the input distribution differs between the training and test stages while the conditional distribution of the output given the input remains unchanged. Most of the existing methods for covariate shift adaptation are two-step approaches, which first calculate the importance weights and then conduct importance-weighted empirical risk minimization. In this paper, we propose a novel one-step approach that jointly learns the predictive model and the associated weights in one optimization by minimizing an upper bound of the test risk. We theoretically analyze the proposed method and provide a generalization error bound. We also empirically demonstrate the effectiveness of the proposed method.

Model identification is a crucial problem in chemical industries. In recent years, there has been increasing interest in learning data-driven models utilizing partial knowledge about the system of interest. Most techniques for model identification do not provide the freedom to incorporate any partial information such as the structure of the model. In this article, we propose model identification techniques that could leverage such partial information to produce better estimates. Specifically, we propose Structural Principal Component Analysis (SPCA) which improvises over existing methods like PCA by utilizing the essential structural information about the model. Most of the existing methods or closely related methods use sparsity constraints which could be computationally expensive. Our proposed method is a wise modification of PCA to utilize structural information. The efficacy of the proposed approach is demonstrated using synthetic and industrial case-studies.

We investigate two causes for adversarial vulnerability in deep neural networks: bad data and (poorly) trained models. When trained with SGD, deep neural networks essentially achieve zero training error, even in the presence of label noise, while also exhibiting good generalization on natural test data, something referred to as benign overfitting [2, 10]. However, these models are vulnerable to adversarial attacks. We identify label noise as one of the causes for adversarial vulnerability, and provide theoretical and empirical evidence in support of this. Surprisingly, we find several instances of label noise in datasets such as MNIST and CIFAR, and that robustly trained models incur training error on some of these, i.e. they don't fit the noise. However, removing noisy labels alone does not suffice to achieve adversarial robustness. Standard training procedures bias neural networks towards learning "simple" classification boundaries, which may be less robust than more complex ones. We observe that adversarial training does produce more complex decision boundaries. We conjecture that in part the need for complex decision boundaries arises from sub-optimal representation learning. By means of simple toy examples, we show theoretically how the choice of representation can drastically affect adversarial robustness.

Conventional bidding strategies for online display ad auction heavily relies on observed performance indicators such as clicks or conversions. A bidding strategy naively pursuing these easily observable metrics, however, fails to optimize the profitability of the advertisers. Rather, the bidding strategy that leads to the maximum revenue is a strategy pursuing the performance lift of showing ads to a specific user. Therefore, it is essential to predict the lift-effect of showing ads to each user on their target variables from observed log data. However, there is a difficulty in predicting the lift-effect, as the training data gathered by a past bidding strategy may have a strong bias towards the winning impressions. In this study, we develop Unbiased Lift-based Bidding System, which maximizes the advertisers' profit by accurately predicting the lift-effect from biased log data. Our system is the first to enable high-performing lift-based bidding strategy by theoretically alleviating the inherent bias in the log. Real-world, large-scale A/B testing successfully demonstrates the superiority and practicability of the proposed system.

Dimensionality reduction is an effective method for learning high-dimensional data, which can provide better understanding of decision boundaries in human-readable low-dimensional subspace. Linear methods, such as principal component analysis and linear discriminant analysis, make it possible to capture the correlation between many variables; however, there is no guarantee that the correlations that are important in predicting data can be captured. Moreover, if the decision boundary has strong nonlinearity, the guarantee becomes increasingly difficult. This problem is exacerbated when the data are matrices or tensors that represent relationships between variables. We propose a learning method that searches for a subspace that maximizes the prediction accuracy while retaining as much of the original data information as possible, even if the prediction model in the subspace has strong nonlinearity. This makes it easier to interpret the mechanism of the group of variables behind the prediction problem that the user wants to know. We show the effectiveness of our method by applying it to various types of data including matrices and tensors.

Providing a small set of promising candidates in place of a single prediction is well-suited for many open-ended classification tasks. Conformal Prediction (CP) is a technique for creating classifiers that produce a valid set of predictions that contains the true answer with arbitrarily high probability. In practice, however, standard CP can suffer from both low predictive and computational efficiency during inference---i.e., the predicted set is both unusably large, and costly to obtain. This is particularly pervasive in the considered setting, where the correct answer is not unique and the number of total possible answers is high. In this work, we develop two simple and complementary techniques for improving both types of efficiencies. First, we relax CP validity to arbitrary criterions of success---allowing our framework to make more efficient predictions while remaining "equivalently correct." Second, we amortize cost by conformalizing prediction cascades, in which we aggressively prune implausible labels early on by using progressively stronger classifiers---while still guaranteeing marginal coverage. We demonstrate the empirical effectiveness of our approach for multiple applications in natural language processing and computational chemistry for drug discovery.

Messenger advertisements (ads) give direct and personal user experience yielding high conversion rates and sales. However, people are skeptical about ads and sometimes perceive them as spam, which eventually leads to a decrease in user satisfaction. Targeted advertising, which serves ads to individuals who may exhibit interest in a particular advertising message, is strongly required. The key to the success of precise user targeting lies in learning the accurate user and ad representation in the embedding space. Most of the previous studies have limited the representation learning in the Euclidean space, but recent studies have suggested hyperbolic manifold learning for the distinct projection of complex network properties emerging from real-world datasets such as social networks, recommender systems, and advertising. We propose a framework that can effectively learn the hierarchical structure in users and ads on the hyperbolic space, and extend to the Multi-Manifold Learning. Our method constructs multiple hyperbolic manifolds with learnable curvatures and maps the representation of user and ad to each manifold. The origin of each manifold is set as the centroid of each user cluster. The user preference for each ad is estimated using the distance between two entities in the hyperbolic space, and the final prediction is determined by aggregating the values calculated from the learned multiple manifolds. We evaluate our method on public benchmark datasets and a large-scale commercial messenger system LINE, and demonstrate its effectiveness through improved performance.