One important approach to learning Bayesian networks (BNs) from data uses a scoring metric to evaluate the fitness of any given candidate network for the data base, and applies a search procedure to explore the set of candidate networks. The most usual search methods are greedy hill climbing, either deterministic or stochastic, although other techniques have also been used. In this paper we propose a new algorithm for learning BNs based on a recently introduced metaheuristic, which has been successfully applied to solve a variety of combinatorial optimization problems: ant colony optimization (ACO). We describe all the elements necessary to tackle our learning problem using this metaheuristic, and experimentally compare the performance of our ACO-based algorithm with other algorithms used in the literature. The experimental work is carried out using three different domains: ALARM, INSURANCE and BOBLO.

We present a neurosymbolic approach to the lifelong learning of algorithmic tasks that mix perception and procedural reasoning. Reusing highlevel concepts across domains and learning complex procedures are two key challenges in lifelong learning. We show that a combination of gradientbased learning and symbolic program synthesis can be a more effective response to these challenges than purely neural methods. Concretely, our approach, called HOUDINI, represents neural networks as strongly typed, end-to-end differentiable functional programs that use symbolic higher-order combinators to compose a library of neural functions. Our learning algorithm consists of: (1) a program synthesizer that performs a type-directed search over programs in this language, and decides on the library functions that should be reused and the architectures that should be used to combine them; and (2) a neural module that trains synthesized programs using stochastic gradient descent. We evaluate our approach on three algorithmic tasks. Our experiments show that our type-directed search technique is able to significantly prune the search space of programs, and that the overall approach transfers high-level concepts more effectively than monolithic neural networks as well as traditional transfer learning.

Let's make one thing clear: one year isn't going to fix decades of gender discrimination in computer science and all the problems associated with it. Recent diversity reports show that women still make up only 20 percent of engineers at Google and Facebook, and an even lower proportion at Uber. But after the parade of awful news about the treatment of female engineers in 2017--sexual harassment in Silicon Valley and a Google engineer sending out a memo to his coworkers arguing that women are biologically less adept at programming, just to name a couple--there is actually reason to believe that things are looking up for 2018, especially when it comes to AI. At first glance, AI would seem among least likely areas of programming to be friendly to women. Writing in Fast Company recently, Hanna Wallach, an AI researcher and cofounder of the Women in Machine Learning Conference, said that only 13.5 percent of those working in machine learning are female.

Welcome to Invisibilia Season 4! The NPR program and podcast explores the invisible forces that shape human behavior, and we here at Shots are joining in to probe the science of why we act the way we do. On paper, Shon Hopwood's life doesn't make a lot of sense, not even to him. "I don't have a great excuse as to why I did these things. And everybody always wants that," he tells me. "It closes the circle for people. But that's not really how it happened."

In this webinar with Paul Clarke where we will examine how artificial intelligence and robotics will affect our education system and curriculum in the coming years. How will this impact on schools and higher education. How will affect the curriculum and how will we have enough teachers that are acquainted with teaching in this new digital world where Artificial intelligence will play a major part? Paul Clarke joined Ocado in 2006. In his current role, Paul heads up Ocado Technology, whose 950 software engineers and other IT specialists are responsible for building all the software and IT infrastructure that powers Ocado, and now Morrisons' online grocery business too.

When discussions arise around a digital transformation journey, the focus is generally put on the technology. After all, in theory, it is technologies like machine learning, artificial intelligence, automation and additive manufacturing that drive the efficiencies and speed companies long for. But digital transformation is not all about the technology. Digital transformation shifts the mindset of how organizations create value for their end-customers, and none of that would be possible without educated and agile employees. The technology is great, but companies need to ensure they are educating and empowering employees along the way.

There are many machine learning frameworks. Given that each takes much time to learn, and given that some have a wider user base than others, which one should you use? Here we look briefly at some of the major ones. In picking a tool, you need to ask what is your goal: machine learning or deep learning? Deep learning has come to mean using neural networks to do, for the most part it seems, image recognition.

The investment is part of an AI strategy laid out by President Emmanuel Macron at the elite College de France research institute in Paris, the French presidency said. The goal is to make better use of the French higher education system that trains computer engineers and mathematicians only to see them leave for jobs at top U.S. tech companies. Some of them have secured high-level positions at Alphabet, the parent company of search engine Google, and Facebook, which opened an AI research centre in Paris in 2015. Macron's AI plan was inspired by a government-commissioned report by Cedric Villani, the self-styled "Lady Gaga of Mathematics" and winner of the mathematics equivalent of the Nobel Prize. Villani, who is also a lawmaker in Macron's party, said in the report the brain drain to Silicon Valley companies showed the excellence of French schools.

This class is a great overview of the types of machine-learning models, and some of the history of how those models came into use. The fundamental explanations of complex ideas are generally excellent and very clear, but the practical equations that are necessary for implementations are difficult to understand for someone like me who is not used to reading abstract mathematical equations. Examples of equations that are worked out with explicit values are few and far between and this doesn't help. This makes the programming assignments exceptionally difficult even though the code they require is simple. Also, the amount of time required for this course is enormous, easily 10x what is predicted when there is a programming assignment.

Online learning algorithms have a wide variety of applications in large scale machine learning problems due to their low computational and memory requirements. However, standard online learning methods still suffer some issues such as lower convergence rates and limited capability to select features or to recover the true features. In this paper, we present a novel framework for online learning based on running averages and introduce a series of online versions of some popular existing offline algorithms such as Adaptive Lasso, Elastic Net and Feature Selection with Annealing. We prove the equivalence between our online methods and their offline counterparts and give theoretical feature selection and convergence guarantees for some of them. In contrast to the existing online methods, the proposed methods can extract model with any desired sparsity level at any time. Numerical experiments indicate that our new methods enjoy high feature selection accuracy and a fast convergence rate, compared with standard stochastic algorithms and offline learning algorithms. We also present some applications to large datasets where again the proposed framework shows competitive results compared to popular online and offline algorithms.