There are situations in which we would like to train a neural network to perform a range of tasks. This is usually possible if we can train the network on all tasks simultane ously. The problem is harder if we would like to train the network sequentially, one task after anot her. The na ive approach of training a trained neural network on a new task via gradient descen t leads to a phenomenon known as catastrophic forgetting: the network's performance in previous ly learned tasks rapidly deteriorates as soon as we start training on a new task. Kirkpatrick et al. [2017] propose a novel algorithm, elastic weight co nslidation (EWC), to address this problem, while maintaining the simplicity of relying on backpropagat ion and stochastic gradient descent as the main algorithmic workhorses. The authors observe that catastrophic forgetting would not happen if the network's parameters were learnt in a Bayesian fa shion: instead of obtaining single estimate of parameters θ via gradient descent, we calculate the Bayesian posterior distribut ion p( θ D

Often the challenge associated with tasks like fraud and spam detection[1] is the lack of all likely patterns needed to train suitable supervised learning models. In order to overcome this limitation, such tasks are attempted as outlier or anomaly detection tasks. We also hypothesize that out- liers have behavioral patterns that change over time. Limited data and continuously changing patterns makes learning significantly difficult. In this work we are proposing an approach that detects outliers in large data sets by relying on data points that are consistent. The primary contribution of this work is that it will quickly help retrieve samples for both consistent and non-outlier data sets and is also mindful of new outlier patterns. No prior knowledge of each set is required to extract the samples. The method consists of two phases, in the first phase, consistent data points (non- outliers) are retrieved by an ensemble method of unsupervised clustering techniques and in the second phase a one class classifier trained on the consistent data point set is ap- plied on the remaining sample set to identify the outliers. The approach is tested on three publicly available data sets and the performance scores are competitive.

Principal component analysis (PCA) is largely adopted for chemical process monitoring and numerous PCA-based systems have been developed to solve various fault detection and diagnosis problems. Since PCA-based methods assume that the monitored process is linear, nonlinear PCA models, such as autoencoder models and kernel principal component analysis (KPCA), has been proposed and applied to nonlinear process monitoring. However, KPCA-based methods need to perform eigen-decomposition (ED) on the kernel Gram matrix whose dimensions depend on the number of training data. Moreover, prefixed kernel parameters cannot be most effective for different faults which may need different parameters to maximize their respective detection performances. Autoencoder models lack the consideration of orthogonal constraints which is crucial for PCA-based algorithms. To address these problems, this paper proposes a novel nonlinear method, called neural component analysis (NCA), which intends to train a feedforward neural work with orthogonal constraints such as those used in PCA. NCA can adaptively learn its parameters through backpropagation and the dimensionality of the nonlinear features has no relationship with the number of training samples. Extensive experimental results on the Tennessee Eastman (TE) benchmark process show the superiority of NCA in terms of missed detection rate (MDR) and false alarm rate (FAR). The source code of NCA can be found in https://github.com/haitaozhao/Neural-Component-Analysis.git.

Generative adversarial networks (GANs) are innovative techniques for learning generative models of complex data distributions from samples. Despite remarkable recent improvements in generating realistic images, one of their major shortcomings is the fact that in practice, they tend to produce samples with little diversity, even when trained on diverse datasets. This phenomenon, known as mode collapse, has been the main focus of several recent advances in GANs. Yet there is little understanding of why mode collapse happens and why existing approaches are able to mitigate mode collapse. We propose a principled approach to handling mode collapse, which we call packing. The main idea is to modify the discriminator to make decisions based on multiple samples from the same class, either real or artificially generated. We borrow analysis tools from binary hypothesis testing---in particular the seminal result of Blackwell [Bla53]---to prove a fundamental connection between packing and mode collapse. We show that packing naturally penalizes generators with mode collapse, thereby favoring generator distributions with less mode collapse during the training process. Numerical experiments on benchmark datasets suggests that packing provides significant improvements in practice as well.

The construction of Mapper has emerged in the last decade as a powerful and effective topological data analysis tool that approximates and generalizes other topological summaries, such as the Reeb graph, the contour tree, split, and joint trees. In this paper we study the parallel analysis of the construction of Mapper. We give a provably correct algorithm to distribute Mapper on a set of processors and discuss the performance results that compare our approach to a reference sequential Mapper implementation. We report the performance experiments that demonstrate the efficiency of our method.

Inference, prediction and control of complex dynamical systems from time series is important in many areas, including financial markets, power grid management, climate and weather modeling, or molecular dynamics. The analysis of such highly nonlinear dynamical systems is facilitated by the fact that we can often find a (generally nonlinear) transformation of the system coordinates to features in which the dynamics can be excellently approximated by a linear Markovian model. Moreover, the large number of system variables often change collectively on large time- and length-scales, facilitating a low-dimensional analysis in feature space. In this paper, we introduce a variational approach for Markov processes (VAMP) that allows us to find optimal feature mappings and optimal Markovian models of the dynamics from given time series data. The key insight is that the best linear model can be obtained from the top singular components of the Koopman operator. This leads to the definition of a family of score functions called VAMP-r which can be calculated from data, and can be employed to optimize a Markovian model. In addition, based on the relationship between the variational scores and approximation errors of Koopman operators, we propose a new VAMP-E score, which can be applied to cross-validation for hyper-parameter optimization and model selection in VAMP. VAMP is valid for both reversible and nonreversible processes and for stationary and non-stationary processes or realizations.

Finding the reduced-dimensional structure is critical to understanding complex networks. Existing approaches such as spectral clustering are applicable only when the full network is explicitly observed. In this paper, we focus on the online factorization and partition of implicit large-scale networks based on observations from an associated random walk. We formulate this into a nonconvex stochastic factorization problem and propose an efficient and scalable stochastic generalized Hebbian algorithm. The algorithm is able to process dependent state-transition data dynamically generated by the underlying network and learn a low-dimensional representation for each vertex. By applying a diffusion approximation analysis, we show that the continuous-time limiting process of the stochastic algorithm converges globally to the "principal components" of the Markov chain and achieves a nearly optimal sample complexity. Once given the learned low-dimensional representations, we further apply clustering techniques to recover the network partition. We show that when the associated Markov process is lumpable, one can recover the partition exactly with high probability. We apply the proposed approach to model the traffic flow of Manhattan as city-wide random walks. By using our algorithm to analyze the taxi trip data, we discover a latent partition of the Manhattan city that closely matches the traffic dynamics.

Many radiological studies can reveal the presence of several co-existing abnormalities, each one represented by a distinct visual pattern. In this article we address the problem of learning a distance metric for plain radiographs that captures a notion of "radiological similarity": two chest radiographs are considered to be similar if they share similar abnormalities. Deep convolutional neural networks (DCNs) are used to learn a low-dimensional embedding for the radiographs that is equipped with the desired metric. Two loss functions are proposed to deal with multi-labelled images and potentially noisy labels. We report on a large-scale study involving over 745,000 chest radiographs whose labels were automatically extracted from free-text radiological reports through a natural language processing system. Using 4,500 validated exams, we demonstrate that the methodology performs satisfactorily on clustering and image retrieval tasks. Remarkably, the learned metric separates normal exams from those having radiological abnormalities.

One intuitive way to make forecasts would be to refer to recent time points. Today's stock prices would likely be more similar to yesterday's prices than those from five years ago. Hence, we would give more weight to recent than to older prices in predicting today's price. These correlations between past and present values demonstrate temporal dependence, which forms the basis of a popular time series analysis technique called ARIMA (Autoregressive Integrated Moving Average). ARIMA accounts for both seasonal variability and one-off'shocks' in the past to make future predictions.

What we call "Machine Learning" is none other than the meeting of statistics and the incredible computation power available today (in terms of memory, CPUs, GPUs). This domain has become increasingly visible important because of the digital revolution of companies leading to the production of massive data of different forms and types, at ever increasing rates: Big Data. On a purely mathematical level most of the algorithms used today are already several decades old. In this article I will explain the underlying logic of 8 machine learning algorithms in the simplest possible terms. Assigning a class / category to each of the observations in a dataset is called classification. It is done a posteriori, once the data is recovered.