For robots to navigate the world, they need to be able to make reasonable assumptions about their surroundings and what might happen during a sequence of events. One way that humans come to learn these things is through sound. For infants, poking and prodding objects is not just fun; some studies suggest that it's actually how they develop an intuitive theory of physics. Could it be that we can get machines to learn the same way? Researchers from MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) have demonstrated an algorithm that has effectively learned how to predict sound: When shown a silent video clip of an object being hit, the algorithm can produce a sound for the hit that is realistic enough to fool human viewers.

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

Ensembles can give you a boost in accuracy on your dataset. In this post you will discover how you can create some of the most powerful types of ensembles in Python using scikit-learn. This case study will step you through Boosting, Bagging and Majority Voting and show you how you can continue to ratchet up the accuracy of the models on your own datasets. Ensemble Machine Learning Algorithms in Python with scikit-learn Photo by The United States Army Band, some rights reserved. It assumes you are generally familiar with machine learning algorithms and ensemble methods and that you are looking for information on how to create ensembles in Python.

It is far from being the apocalyptic extermination of the human race depicted in films like the Terminator, but it seems the rise of the machines has begun. Robots are starting to perform a growing number jobs of that have traditionally been filled by humans – such as security patrols, courier deliveries and room service in hotels. These include five-feet tall security bots that are patrolling shopping centres and office buildings in California, using a range of sensors to look for anything suspicious. Robots are starting to perform roles that have traditionally been performed by humans. While there are many who fear robots are on the verge of stealing our jobs, it seems they have a weak spot - flat packed furniture.

Mazes have been used for a long time in research, particularly in testing memory. Historically, the subjects involved in these tests have mostly been mice and rats, but now a completely new kind of intelligence is getting put through its paces. A group of US researchers is testing artificial intelligence machines using mazes in Minecraft to see which kinds of systems are best at solving them. One example of a simple task was to find the red cube in a maze. Junhyuk Oh and his team created a set of mazes and set their AI algorithms increasingly complex tasks to complete within the mazes.

Within a decade marketers will have to worry more about the algorithm than emotion Chris Stephenson, chief strategist for media agency PHD, has predicted. The bold pronouncement was made during the international keynote last week at the Mumbrella360 conference with Stephenson warning that humans are not far away from a future where consumers will have virtual personal assistants (VPAs) who will make purchasing decisions for them. "The AI (artificial intelligence) train is coming," Stephenson said. "As marketers, as media people this isn't something that will just vaguely enter our world. Artificial intelligence is going to come hurtling through our yard, it is going to change our industry, it is going to change our what we do and how we do it. "It is going to have massive significant impacts on our industry." Stephenson cited the rapid developments in the AI space in recent years and argued that these developments would soon flow to consumers. "Where we are is one minute to midnight on AI eve," he said. "The last few decades have see various technologies and developments that have got us to this moment where we are on the edge of an AI future.

Bottom line: Machine learning is providing the needed algorithms, applications, and frameworks to bring greater predictive accuracy and value to enterprises' data, leading to diverse company-wide strategies succeeding faster and more profitably than before. The good news for businesses is that all the data they have been saving for years can now be turned into a competitive advantage and lead to strategic goals being accomplished. Revenue teams are using machine learning to optimize promotions, compensation and rebates drive the desired behavior across selling channels. Predicting propensity to buy across all channels, making personalized recommendations to customers, forecasting long-term customer loyalty and anticipating potential credit risks of suppliers and buyers are Figure 1 provides an overview of machine learning applications by industry. Unlike advanced analytics techniques that seek out causality first, machine learning techniques are designed to seek out opportunities to optimize decisions based on the predictive value of large-scale data sets.

The American Diabetes Association and IBM Watson Health will partner on a long-term collaboration to bring together the cognitive computing power of Watson and the ADA's vast clinical and research data. Kevin L. Hagen, ADA CEO, said the ADA and IBM Watson Health will work to build a first-of-its-kind diabetes advisor for patients and caregivers, as well as develop Watson-powered solutions to optimize clinical, research and lifestyle decisions. "The goal of this endeavor is to develop and introduce cognitive technologies that support clinicians, researchers and people living with diabetes so that we can accelerate discoveries, personalize treatment and care, and improve the lives of all people affected by diabetes," Hagen said while announcing the partnership. "We also see potential to address social determinants of health." Unlike traditional computing systems that are programmed, systems like IBM Watson learn at scale, explained IBM General Manager David Kenny.

In this work, we consider compressed sensing reconstruction from $M$ measurements of $K$-sparse structured signals which do not possess a writable correlation model. Assuming that a generative statistical model, such as a Boltzmann machine, can be trained in an unsupervised manner on example signals, we demonstrate how this signal model can be used within a Bayesian framework of signal reconstruction. By deriving a message-passing inference for general distribution restricted Boltzmann machines, we are able to integrate these inferred signal models into approximate message passing for compressed sensing reconstruction. Finally, we show for the MNIST dataset that this approach can be very effective, even for $M < K$.

The estimation of class prevalence, i.e., the fraction of a population that belongs to a certain class, is a very useful tool in data analytics and learning, and finds applications in many domains such as sentiment analysis, epidemiology, etc. For example, in sentiment analysis, the objective is often not to estimate whether a specific text conveys a positive or a negative sentiment, but rather estimate the overall distribution of positive and negative sentiments during an event window. A popular way of performing the above task, often dubbed quantification, is to use supervised learning to train a prevalence estimator from labeled data. Contemporary literature cites several performance measures used to measure the success of such prevalence estimators. In this paper we propose the first online stochastic algorithms for directly optimizing these quantification-specific performance measures. We also provide algorithms that optimize hybrid performance measures that seek to balance quantification and classification performance. Our algorithms present a significant advancement in the theory of multivariate optimization and we show, by a rigorous theoretical analysis, that they exhibit optimal convergence. We also report extensive experiments on benchmark and real data sets which demonstrate that our methods significantly outperform existing optimization techniques used for these performance measures.