Ji\v{r}\'i Matou\v{s}ek (1963-2015) had many breakthrough contributions in mathematics and algorithm design. His milestone results are not only profound but also elegant. By going beyond the original objects --- such as Euclidean spaces or linear programs --- Jirka found the essence of the challenging mathematical/algorithmic problems as well as beautiful solutions that were natural to him, but were surprising discoveries to the field. In this short exploration article, I will first share with readers my initial encounter with Jirka and discuss one of his fundamental geometric results from the early 1990s. In the age of social and information networks, I will then turn the discussion from geometric structures to network structures, attempting to take a humble step towards the holy grail of network science, that is to understand the network essence that underlies the observed sparse-and-multifaceted network data. I will discuss a simple result which summarizes some basic algebraic properties of personalized PageRank matrices. Unlike the traditional transitive closure of binary relations, the personalized PageRank matrices take "accumulated Markovian closure" of network data. Some of these algebraic properties are known in various contexts. But I hope featuring them together in a broader context will help to illustrate the desirable properties of this Markovian completion of networks, and motivate systematic developments of a network theory for understanding vast and ubiquitous multifaceted network data.

It has been postulated that a good representation is one that disentangles the underlying explanatory factors of variation. However, it remains an open question what kind of training framework could potentially achieve that. Whereas most previous work focuses on the static setting (e.g., with images), we postulate that some of the causal factors could be discovered if the learner is allowed to interact with its environment. The agent can experiment with different actions and observe their effects. More specifically, we hypothesize that some of these factors correspond to aspects of the environment which are independently controllable, i.e., that there exists a policy and a learnable feature for each such aspect of the environment, such that this policy can yield changes in that feature with minimal changes to other features that explain the statistical variations in the observed data. We propose a specific objective function to find such factors and verify experimentally that it can indeed disentangle independently controllable aspects of the environment without any extrinsic reward signal.

This article was written by Roopam Upadhyay. Roopam is a seasoned professional of advanced analytics with more than a decade of experience in statistical modeling, data science, predictive analytics, optimization, & business consulting. They learn the same way as humans. Humans learn from experience and so do machines. For machines, experience is in the form of data.

I recently curated this dataset to explore some algorithmic approximation of the categories that make up our news, a thing that at different times I have both read and created. If you had tens of thousands of articles from a spread of outlets that seem more or less representative of our national news landscape and you turned them into structured data, and you put a gun to that data's head and coerced it into groups, what would those groups be? I decided the best balance of simplicity and efficacy would be to use unsupervised clustering methods and let the data sort itself, however crudely (and categories, no matter what algorithm they're derived from, will almost always be crude, as there's no reason the media can't be infinitesimally taxonomized). For a variety of reasons (local memory constraints, ability, recommendations from those more learned), I chose to run a bag-of-words through KMeans -- in other words, if every word becomes its own dimension and each article a single datapoint, what clusters of articles will form? If those words already bore you and you're itching to skip to the "so what" and/or don't care about code, scroll down until you see bold letters telling you not to. The code is here if anyone wants to peer-review this and tell me if/where I screwed up and/or give me suggestions.

Do you ever feel overwhelmed by an unending stream of information? It can seem like a barrage of new email and text messages demands constant attention, and there are also phone calls to pick up, articles to read, and knocks on the door to answer. Putting these pieces together to keep track of what is important can be a real challenge. The same information overload is a concern in many computational settings. Telecommunications companies, for example, want to keep track of the activity on their networks, to identify overall network health and spot anomalies or changes in behavior. Yet, the scale of events occurring is huge: many millions of network events per hour, per network element. While new technologies allow the scale and granularity of events being monitored to increase by orders of magnitude, the capacity of computing elements (processors, memory, and disks) to make sense of these is barely increasing. Even on a small scale, the amount of information may be too large to store in an impoverished setting (say, an embedded device) or to keep conveniently in fast storage. In response to this challenge, the model of streaming data processing has grown in popularity. The aim is no longer to capture, store, and index every minute event, but rather to process each observation quickly in order to create a summary of the current state. Following its processing, an event is dropped and is no longer accessible. The summary that is retained is often referred to as a sketch of the data. Coping with the vast scale of information means making compromises: The description of the world is approximate rather than exact; the nature of queries to be answered must be decided in advance rather than after the fact; and some questions are now insoluble. The ability to process vast quantities of data at blinding speeds with modest resources, however, can more than make up for these limitations.

Machine learning is the subfield of computer science that gives computers the ability to learn without being explicitly programmed. It explores the study and construction of algorithms that can learn from and make predictions on data. The R language is widely used among statisticians and data miners to develop statistical software and data analysis. In this course, you will work through various examples on advanced algorithms, and focus a bit more on some visualization options. We'll start by showing you how to use random forest to predict what type of insurance a patient has based on their treatment and you will get an overview of how to use random forest/decision tree and examine the model.

We present ease.ml, a declarative machine learning service platform we built to support more than ten research groups outside the computer science departments at ETH Zurich for their machine learning needs. With ease.ml, a user defines the high-level schema of a machine learning application and submits the task via a Web interface. The system automatically deals with the rest, such as model selection and data movement. In this paper, we describe the ease.ml architecture and focus on a novel technical problem introduced by ease.ml regarding resource allocation. We ask, as a "service provider" that manages a shared cluster of machines among all our users running machine learning workloads, what is the resource allocation strategy that maximizes the global satisfaction of all our users? Resource allocation is a critical yet subtle issue in this multi-tenant scenario, as we have to balance between efficiency and fairness. We first formalize the problem that we call multi-tenant model selection, aiming for minimizing the total regret of all users running automatic model selection tasks. We then develop a novel algorithm that combines multi-armed bandits with Bayesian optimization and prove a regret bound under the multi-tenant setting. Finally, we report our evaluation of ease.ml on synthetic data and on one service we are providing to our users, namely, image classification with deep neural networks. Our experimental evaluation results show that our proposed solution can be up to 9.8x faster in achieving the same global quality for all users as the two popular heuristics used by our users before ease.ml.

We comment on the fact that gradient ascent for logistic regression has a connection with the perceptron learning algorithm. Logistic learning is the "soft" variant of perceptron learning.

Short Video Abstract Prediction of disease onset from patient survey and lifestyle data is quickly becoming an important tool for diagnosing a disease before it progresses. In this study, data from the National Health and Nutrition Examination Survey (NHANES) questionnaire is used to predict the onset of type II diabetes. An ensemble model using the output of five classification algorithms was developed to predict the onset on diabetes based on 16 features. The ensemble model had an AUC of 0.834 indicating high performance.

The spectral gap $\gamma$ of a finite, ergodic, and reversible Markov chain is an important parameter measuring the asymptotic rate of convergence. In applications, the transition matrix $P$ may be unknown, yet one sample of the chain up to a fixed time $n$ may be observed. We consider here the problem of estimating $\gamma$ from this data. Let $\pi$ be the stationary distribution of $P$, and $\pi_\star = \min_x \pi(x)$. We show that if $n = \tilde{O}\bigl(\frac{1}{\gamma \pi_\star}\bigr)$, then $\gamma$ can be estimated to within multiplicative constants with high probability. When $\pi$ is uniform on $d$ states, this matches (up to logarithmic correction) a lower bound of $\tilde{\Omega}\bigl(\frac{d}{\gamma}\bigr)$ steps required for precise estimation of $\gamma$. Moreover, we provide the first procedure for computing a fully data-dependent interval, from a single finite-length trajectory of the chain, that traps the mixing time $t_{\text{mix}}$ of the chain at a prescribed confidence level. The interval does not require the knowledge of any parameters of the chain. This stands in contrast to previous approaches, which either only provide point estimates, or require a reset mechanism, or additional prior knowledge. The interval is constructed around the relaxation time $t_{\text{relax}} = 1/\gamma$, which is strongly related to the mixing time, and the width of the interval converges to zero roughly at a $1/\sqrt{n}$ rate, where $n$ is the length of the sample path.