We introduce an unsupervised discriminative model for the task of retrieving experts in online document collections. We exclusively employ textual evidence and avoid explicit feature engineering by learning distributed word representations in an unsupervised way. We compare our model to state-of-the-art unsupervised statistical vector space and probabilistic generative approaches. Our proposed log-linear model achieves the retrieval performance levels of state-of-the-art document-centric methods with the low inference cost of so-called profile-centric approaches. It yields a statistically significant improved ranking over vector space and generative models in most cases, matching the performance of supervised methods on various benchmarks. That is, by using solely text we can do as well as methods that work with external evidence and/or relevance feedback. A contrastive analysis of rankings produced by discriminative and generative approaches shows that they have complementary strengths due to the ability of the unsupervised discriminative model to perform semantic matching.

We consider the problem of nonparametric regression under shape constraints. The main examples include isotonic regression (with respect to any partial order), unimodal/convex regression, additive shape-restricted regression, and constrained single index model. We review some of the theoretical properties of the least squares estimator (LSE) in these problems, emphasizing on the adaptive nature of the LSE. In particular, we study the risk behavior of the LSE, and its pointwise limiting distribution theory, with special emphasis to isotonic regression. We survey various methods for constructing pointwise confidence intervals around these shape-restricted functions. We also briefly discuss the computation of the LSE and indicate some open research problems and future directions.

Recently, digital music libraries have been developed and can be plainly accessed. Latest research showed that current organization and retrieval of music tracks based on album information are inefficient. Moreover, they demonstrated that people use emotion tags for music tracks in order to search and retrieve them. In this paper, we discuss separability of a set of emotional labels, proposed in the categorical emotion expression, using Fisher's separation theorem. We determine a set of adjectives to tag music parts: happy, sad, relaxing, exciting, epic and thriller. Temporal, frequency and energy features have been extracted from the music parts. It could be seen that the maximum separability within the extracted features occurs between relaxing and epic music parts. Finally, we have trained a classifier using Support Vector Machines to automatically recognize and generate emotional labels for a music part. Accuracy for recognizing each label has been calculated; where the results show that epic music can be recognized more accurately (77.4%), comparing to the other types of music.

Latent factor models are increasingly popular for modeling multi-relational knowledge graphs. By their vectorial nature, it is not only hard to interpret why this class of models works so well, but also to understand where they fail and how they might be improved. We conduct an experimental survey of state-of-the-art models, not towards a purely comparative end, but as a means to get insight about their inductive abilities. To assess the strengths and weaknesses of each model, we create simple tasks that exhibit first, atomic properties of binary relations, and then, common inter-relational inference through synthetic genealogies. Based on these experimental results, we propose new research directions to improve on existing models.

Differential privacy (DP), ever since its advent, has been a controversial object. On the one hand, it provides strong provable protection of individuals in a data set, on the other hand, it has been heavily criticized for being not practical, partially due to its complete independence to the actual data set it tries to protect. In this paper, we address this issue by a new and more fine-grained notion of differential privacy --- per instance differential privacy (pDP), which captures the privacy of a specific individual with respect to a fixed data set. We show that this is a strict generalization of the standard DP and inherits all its desirable properties, e.g., composition, invariance to side information and closedness to postprocessing, except that they all hold for every instance separately. When the data is drawn from a distribution, we show that per-instance DP implies generalization. Moreover, we provide explicit calculations of the per-instance DP for the output perturbation on a class of smooth learning problems. The result reveals an interesting and intuitive fact that an individual has stronger privacy if he/she has small "leverage score" with respect to the data set and if he/she can be predicted more accurately using the leave-one-out data set. Using the developed techniques, we provide a novel analysis of the One-Posterior-Sample (OPS) estimator and show that when the data set is well-conditioned it provides $(\epsilon,\delta)$-pDP for any target individuals and matches the exact lower bound up to a $1+\tilde{O}(n^{-1}\epsilon^{-2})$ multiplicative factor. We also propose AdaOPS which uses adaptive regularization to achieve the same results with $(\epsilon,\delta)$-DP. Simulation shows several orders-of-magnitude more favorable privacy and utility trade-off when we consider the privacy of only the users in the data set.

The online problem of computing the top eigenvector is fundamental to machine learning. In both adversarial and stochastic settings, previous results (such as matrix multiplicative weight update, follow the regularized leader, follow the compressed leader, block power method) either achieve optimal regret but run slow, or run fast at the expense of loosing a d factor in total regret where d is the matrix dimension. We propose a follow-the-compressed-leader (FTCL) framework which achieves optimal regret without sacrificing the running time. Our idea is to "compress" the matrix strategy to dimension 3 in the adversarial setting, or dimension 1 in the stochastic setting.

The loss or corruption of data segments of considerable duration is a very common issue in data restoration and transmission. In audio applications in particular, the insertion of perceptually pleasing content is very important. A good insertion would prevent audible artifacts and provide a coherent and meaningful signal to the listener who would, optimally, remain unaware that any problem has occurred. This task has recently become known as audio inpainting [1], but has previously been referred to e.g. as audio interpolation [2] or waveform substitution [3]. Audio inpainting aims at reconstructing missing parts of an audio signal. When missing parts have a length no longer than 50ms, sparsity-based techniques can be successful [1], [4], [5].

IEEE International Conference on Data Mining identified 10 algorithms in 2006 using surveys from past winners and voting. This is a list of those algorithms a short description and related python resources. The detailed paper is given here. C4.5 is an algorithm used to generate a decision tree developed by Ross Quinlan. The decision trees generated by C4.5 can be used for classification, and for this reason, C4.5 is often referred to as a statistical classifier.

A common applied statistics task involves building regression models to characterize non-linear relationships between variables. When we write a function that takes continuous values as inputs, we are essentially implying an infinite vector that only returns values (indexed by the inputs) when the function is called upon to do so. To make this notion of a "distribution over functions" more concrete, let's quickly demonstrate how we obtain realizations from a Gaussian process, which result in an evaluation of a function over a set of points. We are going generate realizations sequentially, point by point, using the lovely conditioning property of mutlivariate Gaussian distributions.

In many datasets, different parts of the data may have their own patterns of correlation, a structure that can be modeled as a mixture of local linear correlation models. The task of finding these mixtures is known as correlation clustering. In this work, we propose a linear correlation clustering method for datasets whose features are pre-divided into two views. The method, called Canonical Least Squares (CLS) clustering, is inspired by multi-output regression and Canonical Correlation Analysis. CLS clusters can be interpreted as variations in the regression relationship between the two views. The method is useful for data mining and data interpretation. Its utility is demonstrated on a synthetic dataset and stock market dataset.