Microsoft and Google have announced plans to expand their machine learning capabilities, through acquisition and new research offices respectively, reports Telecoms.com.

Based in Google Research offices in Zurich, Switzerland, the new group will focus on three key areas of artificial intelligence: machine intelligence, machine perception, and natural language processing and understanding, according to a blog post by Emmanuel Mogenet, head of Google Research for Europe. It will research ways to improve machine-learning infrastructure and enable the technology for practical use, for instance. Researchers will also work closely with linguists to advance natural language understanding, Mogenet said. Zurich, meanwhile, is home to Google's largest engineering office outside the U.S. Researchers there developed the engine that powers Knowledge Graph as well as the conversation engine that powers the Google Assistant in its Allo messaging app. Katherine Noyes has been an ardent geek ever since she first conquered Pyramid of Doom on an ... Full investment news article is here.

A computer algorithm that can identify patterns in the social media activity of Islamic State supporters could provide clues about where terrorist attacks are likely to occur. Scientists have found they can spot distinct behavioural patterns in the interactions between groups on social media, and it could even help them predict'lone wolf' attacks'. Social media has been a key tool for organisations like Isis to help them recruit supporters and coordinate their activities. Scientists at the University of Miami have used equations used in physics and chemistry to track the constantly shifting behaviour of supporters of Isis (Isis flag pictured). While law enforcement authorities have attempted to keep track of Isis members using social media, they have tended to focus on monitoring the posts made by individuals.

Pendo is bringing the attention to what really matters today: data liquidity, which Pendo defines as liberating data by sophisticated matching and search providing for API's and open standards. Pendo is pleased to announce Version 3.1 of the Pendo Data Platform (PDP), surfacing new abilities in the user-friendly web UI incorporating Artificial Intelligence (AI) attacking the battle of aggregation and intelligent machine learning in gathering of Excel spreadsheets. "AI is a magnitude more complex than machine learning," states Pamela Pecs Cytron, CEO of Pendo Systems. When you model the mind you can create systems capable of learning everything about the world. It is a much smaller task, since the world is very large and changes occur behind your back, which means World Models will become obsolete the moment they are made.

Predictive modeling is fundamental to transforming large clinical data sets, or "big clinical data," into actionable knowledge for various healthcare applications. Machine learning is a major predictive modeling approach, but two barriers make its use in healthcare challenging. First, a machine learning tool user must choose an algorithm and assign one or more model parameters called hyper-parameters before model training. The algorithm and hyper-parameter values used typically impact model accuracy by over 40 %, but their selection requires many labor-intensive manual iterations that can be difficult even for computer scientists. Second, many clinical attributes are repeatedly recorded over time, requiring temporal aggregation before predictive modeling can be performed.

Claned Co-founder Vesa Perala believes that instead of attempting to retrofit technology to out-dated educational systems, EdTech start-ups should be helping to write a new rulebook. For the past 3 years, Claned has been in what he describes as semi-stealth mode, focusing on developing a robust artificial intelligence system that uses machine-learning algorithms to map out what factors most impact individual learning. That knowledge, he says, was already out there, because it's something universities routinely do. Over time, tutors build an understanding of how each student learns, yet that data is trapped in a system which simply isn't scalable. Claned set out to solve this by combining these tried-and-tested academic evaluation metrics with machine learning algorithms and Artificial Intelligence.

What do you think of when someone says "AI" or "Artificial Intelligence"? For most of us, it conjures up an image of the future. It doesn't much evoke the here and now. Artificial intelligence is already out of the box. And while it might not be as slick as the movies, it has vast applications in almost every field, from business to medicine, traffic jams to Facebook photos.

Microsoft Corp. has agreed to acquire Wand Labs, a startup whose messaging technology will help upgrade the software giant's efforts in chatbots. Financial terms of the deal weren't disclosed. Earlier this week, Microsoft MSFT, 1.41% agreed to a blockbuster acquisition of LinkedIn Corp. LNKD, 0.02% for 26.2 billion. Redwood City, Calif.-based Wand Labs will join the Bing engineering and platform group at Microsoft. The Bing group has been developing technology that fits into Microsoft's strategy surrounding what Chief Executive Satya Nadella calls "conversation as a platform."

We combine Riemannian geometry with the mean field theory of high dimensional chaos to study the nature of signal propagation in generic, deep neural networks with random weights. Our results reveal an order-to-chaos expressivity phase transition, with networks in the chaotic phase computing nonlinear functions whose global curvature grows exponentially with depth but not width. We prove this generic class of deep random functions cannot be efficiently computed by any shallow network, going beyond prior work restricted to the analysis of single functions. Moreover, we formalize and quantitatively demonstrate the long conjectured idea that deep networks can disentangle highly curved manifolds in input space into flat manifolds in hidden space. Our theoretical analysis of the expressive power of deep networks broadly applies to arbitrary nonlinearities, and provides a quantitative underpinning for previously abstract notions about the geometry of deep functions.

The stochastic linear bandit problem proceeds in rounds where at each round the algorithm selects a vector from a decision set after which it receives a noisy linear loss parameterized by an unknown vector. The goal in such a problem is to minimize the (pseudo) regret which is the difference between the total expected loss of the algorithm and the total expected loss of the best fixed vector in hindsight. In this paper, we consider settings where the unknown parameter has structure, e.g., sparse, group sparse, low-rank, which can be captured by a norm, e.g., $L_1$, $L_{(1,2)}$, nuclear norm. We focus on constructing confidence ellipsoids which contain the unknown parameter across all rounds with high-probability. We show the radius of such ellipsoids depend on the Gaussian width of sets associated with the norm capturing the structure. Such characterization leads to tighter confidence ellipsoids and, therefore, sharper regret bounds compared to bounds in the existing literature which are based on the ambient dimensionality.