The number of component classifiers chosen for an ensemble has a great impact on its prediction ability. In this paper, we use a geometric framework for a priori determining the ensemble size, applicable to most of the existing batch and online ensemble classifiers. There are only a limited number of studies on the ensemble size considering Majority Voting (MV) and Weighted Majority Voting (WMV). Almost all of them are designed for batch-mode, barely addressing online environments. The big data dimensions and resource limitations in terms of time and memory make the determination of the ensemble size crucial, especially for online environments. Our framework proves, for the MV aggregation rule, that the more strong components we can add to the ensemble the more accurate predictions we can achieve. On the other hand, for the WMV aggregation rule, we prove the existence of an ideal number of components equal to the number of class labels, with the premise that components are completely independent of each other and strong enough. While giving the exact definition for a strong and independent classifier in the context of an ensemble is a challenging task, our proposed geometric framework provides a theoretical explanation of diversity and its impact on the accuracy of predictions. We conduct an experimental evaluation with two different scenarios to show the practical value of our theorems.

At 10 public schools in Cincinnati, middle and high school students will have a new app looking out for them this year. When a student from those schools goes to the health clinic for a talk with the staff psychologist, an iPhone app will listen to the conversation and flag those students it considers likely to attempt suicide. There's a dire need for tech that can detect young people who need help. Suicide is the second-leading cause of death for people ages 15 to 24, surpassed only by accidents. The tech, which has been tested in the Cincinnati schools during the past two years, comes from John Pestian, director of the computational medicine lab at Cincinnati Children's Hospital.

Last spring, Katie Rivera's daughter came home from the St. Francis Parish School in Bakersfield with some unusual paperwork. The school was pushing parents to sign their children up for a "unique pilot program" taught entirely online and run by a public school district in Los Angeles County. Each student who enrolled in the Lennox Virtual Academy would get a free Chromebook computer to use at school, with access to online classes. All parents had to do was fill out the forms, authorizing St. Francis to share information about their finances and their children's health with the Lennox School District a hundred miles away. "This partnership is expected to bring many benefits for St. Francis students," Principal Kelli Gruszka wrote to parents.

We hear a lot these days about the "citizen data scientist." Everyone wants to use data science and machine learning to understand their business and automate tasks to improve efficiency. But we have a shortage of people with data science skills, so much so that salaries are high for properly qualified people. To chief data officers, it's an attractive proposition to take people from within their business who understand data and have a strong mathematical background and convert them to data scientists through self-study and online courses. We have a new generation of visual composition framework tools that enable a business user to visually compose pipelines of algorithms, using techniques such as R and Python selectively to solve more complex problems.

When Katharine Rubin has a spare moment on the way to school, she helps a big-name tech company smarten up its artificial intelligence. Rubin, a 22-year-old accounting major at New York City's Baruch College, is part of a growing workforce that spends anywhere from 5 minutes to 40 hours a week increasing the I in AI. Specifically, Rubin and others provide training data for machine learning algorithms, a form of AI that can be taught from experience. For an autonomous car to recognize pedestrians and stop signs, it's typically fed thousands or millions of photos, all hand-labeled. To nail a conversation, a digital assistant needs to be told over and over when it's failed.

Chatbots are about to change the world in more ways than we can imagine. Already, bots around the globe can complete a diverse set of varying tasks. From ordering pizza online to mashing faces together in Project Murphy, chatbots are about to become a normal element in everyday life. As the scope of chatbots becomes broader every day, there are new applications popping up constantly. Education has traditionally been known as a sector where innovation moves slowly.

Department of Informatics, King's College London; osvaldo.simeone@kcl.ac.uk ABSTRACT This monograph aims at providing an introduction to key concepts, algorithms, and theoretical frameworks in machine learning, including supervised and unsupervised learning, statistical learning theory, probabilistic graphical models and approximate inference. The intended readership consists of electrical engineers with a background in probability and linear algebra. The treatment builds on first principles, and organizes the main ideas according to clearly defined categories, such as discriminative and generative models, frequentist and Bayesian approaches, exact and approximate inference, directed and undirected models, and convex and non-convex optimization. The mathematical framework uses information-theoretic measures as a unifying tool. The text offers simple and reproducible numerical examples providing insights into key motivations and conclusions. Rather than providing exhaustive details on the existing myriad solutions in each specific category, for which the reader is referred to textbooks and papers, this monograph is meant as an entry point for an engineer into the literature on machine learning.

Recently, much work has been done on extending the scope of online learning and incremental stochastic optimization algorithms. In this paper we contribute to this effort in two ways: First, based on a new regret decomposition and a generalization of Bregman divergences, we provide a self-contained, modular analysis of the two workhorses of online learning: (general) adaptive versions of Mirror Descent (MD) and the Follow-the-Regularized-Leader (FTRL) algorithms. The analysis is done with extra care so as not to introduce assumptions not needed in the proofs and allows to combine, in a straightforward way, different algorithmic ideas (e.g., adaptivity, optimism, implicit updates) and learning settings (e.g., strongly convex or composite objectives). This way we are able to reprove, extend and refine a large body of the literature, while keeping the proofs concise. The second contribution is a byproduct of this careful analysis: We present algorithms with improved variational bounds for smooth, composite objectives, including a new family of optimistic MD algorithms with only one projection step per round. Furthermore, we provide a simple extension of adaptive regret bounds to practically relevant non-convex problem settings with essentially no extra effort.

Scientists have discovered that the range of emotions humans experience is much wider than previously thought. While it was originally thought we feel just six emotions, researchers at UC Berkeley found 27 distinct human emotions and have displayed them on an interactive map. In addition to happiness, sadness, anger, surprise, fear, and, disgust, they also determined confusion, romance, nostalgia, sexual desire, and others to be distinct emotions. The emotion map the researchers created: In addition to happiness, sadness, anger, surprise, fear, and, disgust, they also determined confusion, romance, nostalgia, sexual desire, and others to be distinct emotions. 'We wanted to shed light on the full palette of emotions that color our inner world,' lead author Alan Cowen said of the study, which was published today in Proceedings of the National Academy of Sciences.

The battle for artificial-intelligence expertise is forcing Apple Inc. AAPL -0.49% to grapple with its famous penchant for secrecy, as tech companies seek to woo talent in a discipline known for its openness. The technology giant this year has been trying to draw attention--but only so much--to its efforts to develop artificial intelligence, or AI, a term that generally describes software that enables computers to learn and improve functions on their own. Apple launched a public blog in July to talk about its work, for example, and has allowed its researchers to speak at several conferences on artificial intelligence, including a TED Talk in April by Tom Gruber, co-creator of Apple's Siri voice assistant, that was posted on YouTube last month. Talking up transparency is unusual for a company whose chief executive, Tim Cook, once joked that it is more secretive than the Central Intelligence Agency. The shift is driven by AI's growing importance in areas like self-driving cars and voice assistants such as Siri.