Online Passive-Aggressive (PA) learning is a class of online margin-based algorithms suitable for a wide range of real-time prediction tasks, including classification and regression. PA algorithms are formulated in terms of deterministic point-estimation problems governed by a set of user-defined hyperparameters: the approach fails to capture model/prediction uncertainty and makes their performance highly sensitive to hyperparameter configurations. In this paper, we introduce a novel PA learning framework for regression that overcomes the above limitations. We contribute a Bayesian state-space interpretation of PA regression, along with a novel online variational inference scheme, that not only produces probabilistic predictions, but also offers the benefit of automatic hyperparameter tuning. Experiments with various real-world data sets show that our approach performs significantly better than a more standard, linear Gaussian state-space model.

This is the author's version which is being continuously updated and not synchronised with the journal version. The final printed version will appear in Briefings in Bioinformatics Abstract Nucleosome positioning is an important process required for proper genome packing and its accessibility to execute the genetic program in a cell-specific, timely manner. In the recent years hundreds of papers have been devoted to the bioinformatics, physics and biology of nucleosome positioning. The purpose of this review is to cover a practical aspect of this field, namely to provide a guide to the multitude of nucleosome positioning resources available online. These include almost 300 experimental datasets of genome-wide nucleosome occupancy profiles determined in different cell types and more than 40 computational tools for the analysis of experimental nucleosome positioning data and prediction of intrinsic nucleosome formation probabilities from the DNA sequence. A manually curated, up to date list of these resources will be maintained at http://generegulation.info. 1 Introduction The nucleosome is the basic unit of chromatin compaction, composed of the histone octamer and 146-147 base pairs (bp) of DNA wrapped around it. Nucleosomes can form at any genomic locations, but some DNA sequences have higher affinity to the histone octamer, mostly due to the differential bending properties of the DNA double helix. In addition, nucleosome positioning is cell type-specific, in a sense that the cells of the same organism sharing the same genome can have different nucleosome locations depending on the cell type and state. Interested readers are directed to a number of recent publications reviewing the biological, physical and bioinformatics aspects of these phenomena, which will be outside of the scope of the current work [1-32]. Here we will omit fundamental scientific questions, and will focus on a very practical aspect of the field: which experimental nucleosome positioning datasets already exist, how to generate your own data, and how to compare these with other experimental datasets and bioinformatically predicted nucleosome positions in a given genome? 1. Available experimental datasets Recent high-throughput genome-wide data with respect to nucleosome positioning come from a number of related techniques, which have in common an idea to cut DNA between nucleosomes and map protected DNA regions. The most frequently used method is MNase-seq (chromatin digestion by micrococcal nuclease followed by deep sequencing) [11, 33-35].

Crowdsourcing utilizes the wisdom of crowds for collective classification via information (e.g., labels of an item) provided by labelers. Current crowdsourcing algorithms are mainly unsupervised methods that are unaware of the quality of crowdsourced data. In this paper, we propose a supervised collective classification algorithm that aims to identify reliable labelers from the training data (e.g., items with known labels). The reliability (i.e., weighting factor) of each labeler is determined via a saddle point algorithm. The results on several crowdsourced data show that supervised methods can achieve better classification accuracy than unsupervised methods, and our proposed method outperforms other algorithms.

We theoretically and experimentally investigate tensor-based regression and classification. Our focus is regularization with various tensor norms, including the overlapped trace norm, the latent trace norm, and the scaled latent trace norm. We first give dual optimization methods using the alternating direction method of multipliers, which is computationally efficient when the number of training samples is moderate. We then theoretically derive an excess risk bound for each tensor norm and clarify their behavior. Finally, we perform extensive experiments using simulated and real data and demonstrate the superiority of tensor-based learning methods over vector- and matrix-based learning methods.

Nonnegative matrix factorization (NMF) has an established reputation as a useful data analysis technique in numerous applications. However, its usage in practical situations is undergoing challenges in recent years. The fundamental factor to this is the increasingly growing size of the datasets available and needed in the information sciences. To address this, in this work we propose to use structured random compression, that is, random projections that exploit the data structure, for two NMF variants: classical and separable. In separable NMF (SNMF) the left factors are a subset of the columns of the input matrix. We present suitable formulations for each problem, dealing with different representative algorithms within each one. We show that the resulting compressed techniques are faster than their uncompressed variants, vastly reduce memory demands, and do not encompass any significant deterioration in performance. The proposed structured random projections for SNMF allow to deal with arbitrarily shaped large matrices, beyond the standard limit of tall-and-skinny matrices, granting access to very efficient computations in this general setting. We accompany the algorithmic presentation with theoretical foundations and numerous and diverse examples, showing the suitability of the proposed approaches.

Department of Computing Imperial College London 180 Queen's Gate London SW7 2RH United Kingdom April 2015 Abstract This paper critically assesses the anti-functionalist stance on consciousness adopted by certain advocates of integrated information theory (IIT), a corollary of which is that human-level artificial intelligence implemented on conventional computing hardware is necessarily not conscious. The critique draws on variations of a well-known gradual neuronal replacement thought experiment, as well as bringing out tensions in IIT's treatment of self-knowledge. The aim, though, is neither to reject IIT outright nor to champion functionalism in particular. Rather, it is suggested that both ideas have something to offer a scientific understanding of consciousness, as long as they are not dressed up as solutions to illusory metaphysical problems. As for human-level AI, we must await its development before we can decide whether or not to ascribe consciousness to it.

Undirected graphical models, or Markov networks, are a popular class of statistical models, used in a wide variety of applications. Popular instances of this class include Gaussian graphical models and Ising models. In many settings, however, it might not be clear which subclass of graphical models to use, particularly for non-Gaussian and non-categorical data. In this paper, we consider a general sub-class of graphical models where the node-wise conditional distributions arise from exponential families. This allows us to derive multivariate graphical model distributions from univariate exponential family distributions, such as the Poisson, negative binomial, and exponential distributions. Our key contributions include a class of M-estimators to fit these graphical model distributions; and rigorous statistical analysis showing that these M-estimators recover the true graphical model structure exactly, with high probability. We provide examples of genomic and proteomic networks learned via instances of our class of graphical models derived from Poisson and exponential distributions.

While stochastic variational inference is relatively well known for scaling inference in Bayesian probabilistic models, related methods also offer ways to circumnavigate the approximation of analytically intractable expectations. The key challenge in either setting is controlling the variance of gradient estimates: recent work has shown that for continuous latent variables, particularly multivariate Gaussians, this can be achieved by using the gradient of the log posterior. In this paper we apply the same idea to gamma distributed latent variables given gamma variational distributions, enabling straightforward "black box" variational inference in models where sparsity and non-negativity are appropriate. We demonstrate the method on a recently proposed gamma process model for network data, as well as a novel sparse factor analysis. We outperform generic sampling algorithms and the approach of using Gaussian variational distributions on transformed variables.

Next generation telecom services will execute on the telecom cloud, which combine the flexibility of today's computing clouds with the service quality of telecom systems. Real-time service assurance will become an integral part in transforming the general and flexible cloud into a robust and highly available cloud that can ensure low latency and agreed service quality to its customers. A service assurance system for telecom services must be able to detect and preferably also predict problems that may violate the agreed service quality. This is a complex task already in legacy systems and will become even more challenging when executing the services in the cloud. Further, the service assurance system must be able to remedy, in real time, these problems once detected. One promising approach to service assurance is based on machine learning, where the service quality and behavior is learned from observations of the system. The ambition is to do automated real-time predictions of the service quality in order to execute mitigation actions in a proactive manner. Machine learning has been used in the past to build prediction models for service quality assurance.

Poyiadjis et al. (2011) show how particle methods can be used to estimate both the score and the observed information matrix for state space models. These methods either suffer from a computational cost that is quadratic in the number of particles, or produce estimates whose variance increases quadratically with the amount of data. This paper introduces an alternative approach for estimating these terms at a computational cost that is linear in the number of particles. The method is derived using a combination of kernel density estimation, to avoid the particle degeneracy that causes the quadratically increasing variance, and Rao-Blackwellisation. Crucially, we show the method is robust to the choice of bandwidth within the kernel density estimation, as it has good asymptotic properties regardless of this choice. Our estimates of the score and observed information matrix can be used within both online and batch procedures for estimating parameters for state space models. Empirical results show improved parameter estimates compared to existing methods at a significantly reduced computational cost. Supplementary materials including code are available.