Search

Within only a few years after the launch of video sharing platforms, viral videos have become a pervasive Internet phenomenon. Yet, notwithstanding growing scholarly interest, the suitability of the viral metaphor seems not to have been studied so far. In this paper, we therefore investigate the attention dynamics of viral videos from the point of view of mathematical epidemiology. We introduce a novel probabilistic model of the progression of infective diseases and use it to analyze time series of YouTube view counts and Google searches. Our results on a data set of almost 800 videos show that their attention dynamics are indeed well accounted for by our epidemic model. In particular, we find that the vast majority of videos considered in this study show very high infection rates.

Sum-Product Networks (SPNs) are a deep probabilistic architecture that up to now has been successfully employed for tractable inference. Here, we extend their scope towards unsupervised representation learning: we encode samples into continuous and categorical embeddings and show that they can also be decoded back into the original input space by leveraging MPE inference. We characterize when this Sum-Product Autoencoding (SPAE) leads to equivalent reconstructions and extend it towards dealing with missing embedding information. Our experimental results on several multi-label classification problems demonstrate that SPAE is competitive with state-of-the-art autoencoder architectures, even if the SPNs were never trained to reconstruct their inputs.

While all kinds of mixed data---from personal data, over panel and scientific data, to public and commercial data---are collected and stored, building probabilistic graphical models for these hybrid domains becomes more difficult. Users spend significant amounts of time in identifying the parametric form of the random variables (Gaussian, Poisson, Logit, etc.) involved and learning the mixed models. To make this difficult task easier, we propose the first trainable probabilistic deep architecture for hybrid domains that features tractable queries. It is based on Sum-Product Networks (SPNs) with piecewise polynomial leaf distributions together with novel nonparametric decomposition and conditioning steps using the Hirschfeld-Gebelein-Renyi Maximum Correlation Coefficient. This relieves the user from deciding a-priori the parametric form of the random variables but is still expressive enough to effectively approximate any distribution and permits efficient learning and inference.Our experiments show that the architecture, called Mixed SPNs, can indeed capture complex distributions across a wide range of hybrid domains.

The learning with privileged information setting has recently attracted a lot of attention within the machine learning community, as it allows the integration of additional knowledge into the training process of a classifier, even when this comes in the form of a data modality that is not available at test time. Here, we show that privileged information can naturally be treated as noise in the latent function of a Gaussian Process classifier (GPC). That is, in contrast to the standard GPC setting, the latent function is not just a nuisance but a feature: it becomes a natural measure of confidence about the training data by modulating the slope of the GPC sigmoid likelihood function. Extensive experiments on public datasets show that the proposed GPC method using privileged noise, called GPC+, improves over a standard GPC without privileged knowledge, and also over the current state-of-the-art SVM-based method, SVM+. Moreover, we show that advanced neural networks and deep learning methods can be compressed as privileged information., 14 pages with figures

Color refinement is a basic algorithmic routine for graph isomorphismtesting and has recently been used for computing graph kernels as well as for lifting belief propagation and linear programming. So far, color refinement has been treated as a combinatorial problem. Instead, we treat it as a nonlinear continuous optimization problem and prove thatit implements a conditional gradient optimizer that can be turned into graph clustering approaches using hashing and truncated power iterations. This shows that color refinement is easy to understand in terms of random walks, easy to implement (matrix-matrix/vector multiplications) and readily parallelizable. We support our theoretical results with experiments on real-world graphs with millions of edges.

In this paper we introduce an approximation of the Weisfeiler-Lehman graph kernel algorithm aimed at improving the computation time of the kernel when applied to Resource Description Framework (RDF) data. Typically, applying graph kernels to RDF is done by extracting subgraphs from a large RDF graph and computing the kernel on this set of subgraphs. In contrast, our algorithm computes the Weisfeiler-Lehman kernel directly on the large RDF graph, but still retains the subgraph information. We show that this algorithm is faster than the regular Weisfeiler-Lehman kernel for RDF data and has at least the same performance. Furthermore, we show that our method has similar or better performance, and is faster, than other recently introduced graph kernels for RDF.

We present an extension of convex-hull non-negative matrix factorization (CH-NMF) which was recently proposed as a large scale variant of convex non-negative matrix factorization or Archetypal Analysis. CH-NMF factorizes a non-negative data matrix V into two non- negative matrix factors V WH such that the columns of W are convex combinations of certain data points so that they are readily interpretable to data analysts. There is, however, no free lunch: imposing convexity constraints on W typically prevents adaptation to intrinsic, low dimensional structures in the data. Alas, in cases where the data is distributed in a non-convex manner or consists of mixtures of lower dimensional convex distributions, the cluster representatives obtained from CH-NMF will be less meaningful. In this paper, we present a hierarchical CH-NMF that automatically adapts to internal structures of a dataset, hence it yields meaningful and interpretable clusters for non-convex datasets. This i s also confirmed by our extensive evaluation on DBLP publication records of 760,000 authors, 4,000,000 images harvested from the web, and 150,000,000 votes on World of Warcraft guilds.

Mobile digital games are dominantly released under the freemium business model, but only a small fraction of the players makes any purchases. The ability to predict who will make a purchase enables optimization of marketing efforts, and tailoring customer relationship management to the specific user's profile. Here this challenge is addressed via two models for predicting purchasing players, using a 100,000 player dataset: 1) A classification model focused on predicting whether a purchase will occur or not. 2) a regression model focused on predicting the number of purchases a user will make. Both models are presented within a decision and regression tree framework for building rules that are actionable by companies. To the best of our knowledge, this is the first study investigating purchase decisions in freemium mobile products from a user behavior perspective and adopting behavior-driven learning approaches to this problem.

Feature selection has been studied in machine learning and data mining for many years, and is a valuable way to improve classification accu- racy while reducing model complexity. Two main classes of feature selection methods - filter and wrapper - discard those features which are not se- lected, and do not consider them in the predictive model. We propose that these unselected features may instead be used as an additional source of in- formation at train time. We describe a strategy called Learning using Unselected Features (LUFe) that allows selected and unselected features to serve different functions in classification. In this frame- work, selected features are used directly to set the decision boundary, and unselected features are utilised in a secondary role, with no additional cost at test time. Our empirical results on 49 textual datasets show that LUFe can improve classification performance in comparison with standard wrapper and filter feature selection.

Pre-symptomatic drought stress prediction is of great relevance in precision plant protection, ultimately helping to meet the challenge of "How to feed a hungry world?". Unfortunately, it also presents unique computational problems in scale and interpretability: it is a temporal, large-scale prediction task, e.g., when monitoring plants over time using hyperspectral imaging, and features are `things' with a `biological' meaning and interpretation and not just mathematical abstractions computable for any data. In this paper we propose Dirichlet-aggregation regression (DAR) to meet the challenge. DAR represents all data by means of convex combinations of only few extreme ones computable in linear time and easy to interpret.Then, it puts a Gaussian process prior on the Dirichlet distributions induced on the simplex spanned by the extremes. The prior can be a function of any observed meta feature such as time, location, type of fertilization, and plant species. We evaluated DAR on two hyperspectral image series of plants over time with about 2 (resp. 5.8) Billion matrix entries. The results demonstrate that DAR can be learned efficiently and predicts stress well before it becomes visible to the human eye.

Internet memes are a pervasive phenomenon on the social Web. They typically consist of viral catch phrases, images, or videos that spread through instant messaging, (micro) blogs, forums, and social networking sites. Due to their popularity and proliferation, Internet memes attract interest in areas as diverse as marketing, sociology, or computer science and have been dubbed a new form of communication or artistic expression. In this paper, we examine the merits of such claims and analyze how collective attention into Internet memes evolves over time. We introduce and discuss statistical models of the dynamics of fads and fit them to meme related time series obtained from Google Trends. Given data as to more than 200 memes, we find that our models provide more accurate descriptions of the dynamics of growth and decline of collective attention to individual Internet memes than previous approaches from the literature. In short, our results suggest that Internet memes are nothing but fads.

Constraint-based clustering leverages user-provided constraints to produce a clustering that matches the user's expectation. In active constraint-based clustering, the algorithm selects the most informative constraints to query in order to produce good clusterings with as few constraints as possible. A major challenge in constraint-based clustering is handling noise: the majority of existing approaches assume that the provided constraints are correct, while that might not be the case. In this paper, we propose a method to identify and correct noisy constraints in active constraint-based clustering. Our approach reasons probabilistically about the correctness of the user’s answers and asks additional constraints to corroborate or correct the suspicious answers. We demonstrate the method’s effectiveness by incorporating it into COBRAS, a state-of-the-art method for active constraint-based clustering. Compared to COBRAS and other active-constraint-based clustering algorithms, the resulting system produces better clusterings in the presence of noise. ispartof: pages:121-136 ispartof: ECML PKDD 2020: Machine Learning and Knowledge Discovery in Databases vol:12458 pages:121-136 ispartof: European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases location:Ghent, Belgium date:14 Sep - 18 Sep 2020 status: published

Anomaly detection focuses on identifying examples in the data that somehow deviate from what is expected or typical. Algorithms for this task usually assign a score to each example that represents how anomalous the example is. Then, a threshold on the scores turns them into concrete predictions. However, each algorithm uses a different approach to assign the scores, which makes them difficult to interpret and can quickly erode a user’s trust in the predictions. This paper introduces an approach for assessing the reliability of any anomaly detector’s example-wise predictions. To do so, we propose a Bayesian approach for converting anomaly scores to probability estimates. This enables the anomaly detector to assign a confidence score to each prediction which captures its uncertainty in that prediction. We theoretically analyze the convergence behaviour of our confidence estimate. Empirically, we demonstrate the effectiveness of the framework in quantifying a detector’s confidence in its predictions on a large benchmark of datasets.

Many machine learning applications are based on data collected from people, such as their tastes and behaviour as well as biological traits and genetic data. Regardless of how important the application might be, one has to make sure individuals' identities or the privacy of the data are not compromised in the analysis. Differential privacy constitutes a powerful framework that prevents breaching of data subject privacy from the output of a computation. Differentially private versions of many important Bayesian inference methods have been proposed, but there is a lack of an efficient unified approach applicable to arbitrary models. In this contribution, we propose a differentially private variational inference method with a very wide applicability. It is built on top of doubly stochastic variational inference, a recent advance which provides a variational solution to a large class of models. We add differential privacy into doubly stochastic variational inference by clipping and perturbing the gradients. The algorithm is made more efficient through privacy amplification from subsampling. We demonstrate the method can reach an accuracy close to non-private level under reasonably strong privacy guarantees, clearly improving over previous sampling-based alternatives especially in the strong privacy regime., Comment: 10 pages, 5 figures

Early water stress recognition is of great relevance in precision plant breeding and production. Hyperspectral imaging sensors can be a valuable tool for early stress detection with high spatio-temporal resolution. They gather large, high dimensional data cubes posing a significant challenge to data analysis. Classical supervised learning algorithms often fail in applied plant sciences due to their need of labelled datasets, which are difficult to obtain. Therefore, new approaches for unsupervised learning of relevant patterns are needed. We apply for the first time a recent matrix factorisation technique, simplex volume maximisation (SiVM), to hyperspectral data. It is an unsupervised classification approach, optimised for fast computation of massive datasets. It allows calculation of how similar each spectrum is to observed typical spectra. This provides the means to express how likely it is that one plant is suffering from stress. The method was tested for drought stress, applied to potted barley plants in a controlled rain-out shelter experiment and to agricultural corn plots subjected to a two factorial field setup altering water and nutrient availability. Both experiments were conducted on the canopy level. SiVM was significantly better than using a combination of established vegetation indices. In the corn plots, SiVM clearly separated the different treatments, even though the effects on leaf and canopy traits were subtle.