Nasa has announced that it has found evidence of flowing water on Mars. Scientists have long speculated that Recurring Slope Lineae -- or dark patches -- on Mars were made up of briny water but the new findings prove that those patches are caused by liquid water, which it has established by finding hydrated salts. Several hundred camped outside the London store in Covent Garden. The 6s will have new features like a vastly improved camera and a pressure-sensitive "3D Touch" display

My team and I attended several legaltech-focused conferences this month and of course, as expected, machine learning and AI were the topics that everyone wanted to discuss in between sessions. Yet interestingly, at the Emerging Legal Technology Forum put on by Legalx, one of the panelists -- Mark Tamminga, leader of innovation initiatives at at Gowling WLG -- was against using these terms in reference to emerging legal technologies. He pointed out that often what is being called AI is really not that at all, and felt that these words were being used as fancy buzzwords that escape the real mechanics of these technologies. As someone deeply involved in the development community here in Toronto, I wholeheartedly agree with his perspective. Many of the conversations occurring in legaltech around what people are calling machine learning are actually algorithmic solutions preprogrammed (that's right, programmed by humans) to do a particular task; nothing that deviates greatly from anything that's already been done many years ago.

Here's the tech news you need to know this Thursday. The move was announced in a letter by Theranos CEO Elizabeth Holmes. Sources cited by Recode say Google does not currently plan to make a bid and was not moving forward with any effort to do so. The company that's been acquired is called Viv Labs. It includes the likes of the Harry Potter series and Sports Illustrated.

A small drone taught itself to judge distances using only one eye during trials aboard the International Space Station, ESA-backed researchers have reported. Although humans can effortlessly estimate distances with a single eye, robots still lack this capability. "It is a mathematical impossibility to extract distances to objects from one single image, if the object has not been encountered before," explains Guido de Croon from Delft University of Technology, one of the investigators. "But if we recognise something to be a car, then we know its physical characteristics, and we can use that information to estimate its distance from us. A similar logic is what we wanted the drone to learn during our experiment."

Google has announced a research group in Switzerland that will be dedicated to the field of machine learning. As its name suggests, machine learning involves systems that can learn things and come up with predictions from sets of data, without being specifically programmed to do so. It's essential to what we currently think of as "artificial intelligence." Google uses machine learning to power its translation engine, its image-recognition systems, and its Inbox "smart reply" feature. Machine learning improves spam recognition in Gmail and helps Google's driverless cars figure out what's going on around them.

COLORADO SPRINGS, CO--(Marketwired - October 05, 2016) - Cherwell Software, a global leader in IT service management (ITSM) solutions, announced today to customers at its annual conference that the Cherwell Service Management platform is utilizing Microsoft Azure Machine Learning to integrate predictive analytics into its ITSM solution. Azure Machine Learning, part of the Microsoft Cortana Intelligence Suite, provides a cloud-based service that enables companies to apply statistical techniques to large amounts of data and leverage analytics to solve problems and create smarter applications. Cherwell customers can now harness the power of Azure Machine Learning to analyze incident and ticket data stored within Cherwell Service Management, and utilize pattern recognition to create algorithms that assist with incident prioritization, triage, and resolution. "Azure Machine Learning and the Cortana Intelligence Suite offer an easy way to add powerful intelligence to existing applications. By integrating Azure Machine Learning with Cherwell Service Management, Cherwell has introduced the next generation of service management," commented Dawson Stoops, VP of Technical Alliances at Cherwell.

"Sandia National Laboratories researchers are drawing inspiration from neurons in the brain, such as these green fluorescent protein-labeled neurons in a mouse neocortex, with the aim of developing neuro-inspired computing systems to reboot computing. "Summary: Researchers explore neural computing to extend Moore's Law. Sandia explores neural computing to extend Moore's Law. Computation is stuck in a rut. The integrated circuits that powered the past 50 years of technological revolution are reaching their physical limits.

In artificial neural networks, learning from data is a computationally demanding task in which a large number of connection weights are iteratively tuned through stochastic-gradient-based heuristic processes over a cost-function. It is not well understood how learning occurs in these systems, in particular how they avoid getting trapped in configurations with poor computational performance. Here we study the difficult case of networks with discrete weights, where the optimization landscape is very rough even for simple architectures, and provide theoretical and numerical evidence of the existence of rare - but extremely dense and accessible - regions of configurations in the network weight space. We define a novel measure, which we call the "robust ensemble" (RE), which suppresses trapping by isolated configurations and amplifies the role of these dense regions. We analytically compute the RE in some exactly solvable models, and also provide a general algorithmic scheme which is straightforward to implement: define a cost-function given by a sum of a finite number of replicas of the original cost-function, with a constraint centering the replicas around a driving assignment. To illustrate this, we derive several powerful new algorithms, ranging from Markov Chains to message passing to gradient descent processes, where the algorithms target the robust dense states, resulting in substantial improvements in performance. The weak dependence on the number of precision bits of the weights leads us to conjecture that very similar reasoning applies to more conventional neural networks. Analogous algorithmic schemes can also be applied to other optimization problems.

To infer multilayer deep representations of high-dimensional discrete and nonnegative real vectors, we propose an augmentable gamma belief network (GBN) that factorizes each of its hidden layers into the product of a sparse connection weight matrix and the nonnegative real hidden units of the next layer. The GBN's hidden layers are jointly trained with an upward-downward Gibbs sampler that solves each layer with the same subroutine. The gamma-negative binomial process combined with a layer-wise training strategy allows inferring the width of each layer given a fixed budget on the width of the first layer. Example results illustrate interesting relationships between the width of the first layer and the inferred network structure, and demonstrate that the GBN can add more layers to improve its performance in both unsupervisedly extracting features and predicting heldout data. For exploratory data analysis, we extract trees and subnetworks from the learned deep network to visualize how the very specific factors discovered at the first hidden layer and the increasingly more general factors discovered at deeper hidden layers are related to each other, and we generate synthetic data by propagating random variables through the deep network from the top hidden layer back to the bottom data layer.

It was shown recently that the $K$ L1-norm principal components (L1-PCs) of a real-valued data matrix $\mathbf X \in \mathbb R^{D \times N}$ ($N$ data samples of $D$ dimensions) can be exactly calculated with cost $\mathcal{O}(2^{NK})$ or, when advantageous, $\mathcal{O}(N^{dK - K + 1})$ where $d=\mathrm{rank}(\mathbf X)$, $K