Last week, we discussed Multi-class SVM loss; specifically, the hinge loss and squared hinge loss functions. A loss function, in the context of Machine Learning and Deep Learning, allows us to quantify how "good" or "bad" a given classification function (also called a "scoring function") is at correctly classifying data points in our dataset. In fact, if you have done previous work in Deep Learning, you have likely heard of this function before -- do the terms Softmax classifier and cross-entropy loss sound familiar? I'll go as far to say that if you do any work in Deep Learning (especially Convolutional Neural Networks) that you'll run into the term "Softmax": it's the final layer at the end of the network that yields your actual probability scores for each class label. To learn more about Softmax classifiers and the cross-entropy loss function, keep reading.

What does "causality" mean, and how can you represent it mathematically? How can you encode causal assumptions, and what bearing do they have on data analysis? These types of questions are at the core of the practice of data science, but deep knowledge about them is surprisingly uncommon. If you analyze data without regard to causality, you open your results up for the possibility of enormous biases. This includes everything from recommendation system results, to post-hoc reports on observational data, to experiments run without proper holdout groups. I've been blogging a lot recently about causality, and wanted to go through some of the material at a more technical level. Recent posts have been aimed at a more general audience. This one will be aimed at practitioners, and will assume a basic working knowledge of math and data analysis. To get the most from this post you should have a reasonable understanding of linear regression and probability (although we'll review a lot of probability). Prior knowledge of graphical models will make some concepts more familiar, but is not required. Judea Pearl, in his book Causality, constantly remarks that until very recently, causality was a concept in search of a language.

The two most extended density-based approaches to clustering are surely mixture model clustering and modal clustering. In the mixture model approach, the density is represented as a mixture and clusters are associated to the different mixture components. In modal clustering, clusters are understood as regions of high density separated from each other by zones of lower density, so that they are closely related to certain regions around the density modes. If the true density is indeed in the assumed class of mixture densities, then mixture model clustering allows to scrutinize more subtle situations than modal clustering. However, when mixture modeling is used in a nonparametric way, taking advantage of the denseness of the sieve of mixture densities to approximate any density, then the correspondence between clusters and mixture components may become questionable. In this paper we introduce two methods to adopt a modal clustering point of view after a mixture model fit. Numerous examples are provided to illustrate that mixture modeling can also be used for clustering in a nonparametric sense, as long as clusters are understood as the domains of attraction of the density modes.

Kernel and Multiple Kernel Canonical Correlation Analysis (CCA) are employed to classify schizophrenic and healthy patients based on their SNPs, DNA Methylation and fMRI data. Kernel and Multiple Kernel CCA are popular methods for finding nonlinear correlations between high-dimensional datasets. Data was gathered from 183 patients, 79 with schizophrenia and 104 healthy controls. Kernel and Multiple Kernel CCA represent new avenues for studying schizophrenia, because, to our knowledge, these methods have not been used on these data before. Classification is performed via k-means clustering on the kernel matrix outputs of the Kernel and Multiple Kernel CCA algorithm. Accuracies of the Kernel and Multiple Kernel CCA classification are compared to that of the regularized linear CCA algorithm classification, and are found to be significantly more accurate. Both algorithms demonstrate maximal accuracies when the combination of DNA methylation and fMRI data are used, and experience lower accuracies when the SNP data are incorporated.

We consider the estimation and inference of sparse graphical models that characterize the dependency structure of high-dimensional tensor-valued data. To facilitate the estimation of the precision matrix corresponding to each way of the tensor, we assume the data follow a tensor normal distribution whose covariance has a Kronecker product structure. A critical challenge in the estimation and inference of this model is the fact that its penalized maximum likelihood estimation involves minimizing a non-convex objective function. To address it, this paper makes two contributions: (i) In spite of the non-convexity of this estimation problem, we prove that an alternating minimization algorithm, which iteratively estimates each sparse precision matrix while fixing the others, attains an estimator with the optimal statistical rate of convergence. Notably, such an estimator achieves estimation consistency with only one tensor sample, which was not observed in the previous work. (ii) We propose a de-biased statistical inference procedure for testing hypotheses on the true support of the sparse precision matrices, and employ it for testing a growing number of hypothesis with false discovery rate (FDR) control. The asymptotic normality of our test statistic and the consistency of FDR control procedure are established. Our theoretical results are further backed up by thorough numerical studies. We implement the methods into a publicly available R package Tlasso.

Markov chain Monte Carlo (MCMC) algorithms are ubiquitous in Bayesian computations. However, they need to access the full data set in order to evaluate the posterior density at every step of the algorithm. This results in a great computational burden in big data applications. In contrast to MCMC methods, Stochastic Gradient MCMC (SGMCMC) algorithms such as the Stochastic Gradient Langevin Dynamics (SGLD) only require access to a batch of the data set at every step. This drastically improves the computational performance and scales well to large data sets. However, the difficulty with SGMCMC algorithms comes from the sensitivity to its parameters which are notoriously difficult to tune. Moreover, the Root Mean Square Error (RMSE) scales as $\mathcal{O}(c^{-\frac{1}{3}})$ as opposed to standard MCMC $\mathcal{O}(c^{-\frac{1}{2}})$ where $c$ is the computational cost. We introduce a new class of Multilevel Stochastic Gradient Markov chain Monte Carlo algorithms that are able to mitigate the problem of tuning the step size and more importantly of recovering the $\mathcal{O}(c^{-\frac{1}{2}})$ convergence of standard Markov Chain Monte Carlo methods without the need to introduce Metropolis-Hasting steps. A further advantage of this new class of algorithms is that it can easily be parallelised over a heterogeneous computer architecture. We illustrate our methodology using Bayesian logistic regression and provide numerical evidence that for a prescribed relative RMSE the computational cost is sublinear in the number of data items.

HERE is an increasing need to handle large multidimensional datasets that cannot efficiently be analyzed or processed using modern day computers. Due to the curse of dimensionality it is urgent to develop mathematical tools which can evaluate information beyond the properties of large matrices [1]. The essential goal is to reduce the dimensionality of multidimensional data, represented by tensors, with a minimal information loss by compressing the original tensor space to a lower-dimensional tensor space, also called the feature space [1]. Tensor decomposition is the most natural tool to enable such compressions [2]. Until recently, tensor compression is merely based on Tucker decomposition (TD) [3], also known as higher-order singular value decomposition (HOSVD) when orthogonality constraints on factor matrices are imposed [4].

Probabilities can be consistently estimated using random forests. It is, however, unclear how random forests should be updated to make predictions for other centers or at different time points. The first method has been proposed by Elkan and may be used for updating any machine learning approach yielding consistent probabilities, so-called probability machines. The second approach is a new strategy specifically developed for random forests. Using the terminal nodes, which represent conditional probabilities, the random forest is first translated to logistic regression models.

Hamiltonian Monte Carlo (HMC) is a popular Markov chain Monte Carlo (MCMC) algorithm that generates proposals for a Metropolis-Hastings algorithm by simulating the dynamics of a Hamiltonian system. However, HMC is sensitive to large time discretizations and performs poorly if there is a mismatch between the spatial geometry of the target distribution and the scales of the momentum distribution. In particular the mass matrix of HMC is hard to tune well. In order to alleviate these problems we propose relativistic Hamiltonian Monte Carlo, a version of HMC based on relativistic dynamics that introduce a maximum velocity on particles. We also derive stochastic gradient versions of the algorithm and show that the resulting algorithms bear interesting relationships to gradient clipping, RMSprop, Adagrad and Adam, popular optimisation methods in deep learning. Based on this, we develop relativistic stochastic gradient descent by taking the zero-temperature limit of relativistic stochastic gradient Hamiltonian Monte Carlo. In experiments we show that the relativistic algorithms perform better than classical Newtonian variants and Adam.

Spatio-temporal data is intrinsically high dimensional, so unsupervised modeling is only feasible if we can exploit structure in the process. When the dynamics are local in both space and time, this structure can be exploited by splitting the global field into many lower-dimensional "light cones". We review light cone decompositions for predictive state reconstruction, introducing three simple light cone algorithms. These methods allow for tractable inference of spatio-temporal data, such as full-frame video. The algorithms make few assumptions on the underlying process yet have good predictive performance and can provide distributions over spatio-temporal data, enabling sophisticated probabilistic inference.