In recent years, deep learning has taken the lead in predictive accuracy in many fields of machine learning, and companies are struggling to keep up with the speed of innovation. Arno Candel demonstrates how successful enterprises can augment simple statistical models with more accurate data-driven models to gain a competitive edge. Arno describes how to build smart applications that include data munging, model training and validation, and real-time production deployment--every step is based on open source code (R, Python, Java, Scala, JavaScript, REST) that runs on distributed platforms including Hadoop, Spark, and standard compute clusters. Arno also presents use cases from verticals including insurance, fraud, churn, fintech, and marketing and offers live demos of smart applications on large real-world datasets in distributed clusters.

A new low-power, "brain-inspired" supercomputing platform based on IBM chip technology will soon start exploring deep learning for the U.S. nuclear program. Lawrence Livermore National Laboratory announced on Tuesday that it has purchased the platform, based on the TrueNorth neurosynaptic chip IBM introduced in 2014. It will use the technology to evaluate machine-learning and deep-learning applications for the National Nuclear Security Administration. The computer will process data with the equivalent of 16 million neurons and 4 billion synapses and consume roughly as much energy as a tablet PC. Also included will be an accompanying ecosystem consisting of a simulator; a programming language; an integrated programming environment; a library of algorithms and applications; firmware; tools for composing neural networks for deep learning; a teaching curriculum; and cloud enablement.

For much of my career I've worked with technologies for handling "structured information" – that is, data that is well-formed, with understood data types and fields, and almost always either generated by a computer of some sort, or coded by people to be easily understood by a computer. This data was, for many years, the most common form of information available to business. Starting about twenty years ago, some new technologies began to emerge which changed things. And today things are very different. Much of the information we create as people – emails, blog posts, web pages, PDF documents, video, audio, and more – is now conveyed or managed by computers.

Models with latent variables have been proposed and investigated for explaining, understanding, or classifying observed data. If a model is a generative model, observed data are modeled to be as if they were generated by latent variables through parameterized probability distributions. Popular criteria for learning generative models include likelihood or posterior probability, which both evaluate the probability of the given observed data or parameters. Another kind of criteria is mutual information. Mutual information has been used to learn nonlinear generative models [14] in which relationships between observed and latent variables are directly evaluated. It has also been used to learn linear encoding (recognition) models [2, 12]. The relationships between observed and latent variables have greater importance in more complex generative models, e.g., deep learning models [6, 9]. In the pre-training of deep belief networks (DBNs), one of the models or techniques of deep learning, posterior samples of latent variables in the lower layer are used as samples of observed variables in the next, higher layer. For successive layer learning to be possible, latent variables should possess properties that enable such learning.

Are there any Recurrent convolutional neural network network implementations out there? I was hoping to implement the text classifier mentioned in this paper.

Editor's Note: Alex Meshkin is the CEO of Flow Health. Flow Health provides longitudinal care plan coordination and chronic care management services built on top of its platform, which is The Operating System for Value-Based CareSM. We have seen incredible progress in machine learning and artificial intelligence (AI) over the past few years, especially through the application of deep learning algorithms. AI systems will get even better as more data is collected, so faster data gathering and better data integration should lead to smarter and more useful AI systems. Recently I described a new class of system that I believe will take form and leverage AI and combine workflow automation to improve how care is delivered -- I termed this: "Intelligent Clinical Decision Automation."

IBM is about to deliver the foundation of a brain-inspired supercomputer to Lawrence Livermore National Laboratory, one of the federal government's top research institutions. The delivery is one small "blade" within a server rack with 16 chips, dubbed TrueNorth, and is modeled after the way the human brain functions. Silicon Valley is awash in optimism about artificial intelligence, largely based on the progress that deep learning neural networks are making in solving big problems. Companies from Google to Nvidia are hoping they'll provide the AI smarts for self-driving cars and other tough problems. It is within this environment that IBM has been pursuing solutions in brain-inspired supercomputers. The main benefit is that such chips may be able to operate at lower frequencies and get much more work done on a much smaller amount of power.

Machine learning company Maluuba, with headquarters in Waterloo, Ontario and a research office in Montreal, has applied an algorithm to the text of J.K. Rowling's bestselling novel Harry Potter and the Philosopher's Stone, along with several hundred other children's stories, to read text in such a way that it can then answer questions afterward. Maluuba has also just announced the opening of an R&D lab in Montreal, staffed by Yoshua Bengio of the Université de Montréal's Montreal Institute for Learning Algorithms (MILA) in partnership with reinforcement learning expert Richard Sutton from the Alberta Innovates Centre for Machine Learning, to make advances in the fields of Natural Language Understanding (NLU) and artificial intelligence (AI). Taking a deep learning approach, Maluuba trained its algorithm to approach the Harry Potter text from several levels of textual abstraction, word, sentence, paragraph, etc. And while a certain contingent of tech utopians may very well look at Maluuba's case study as the smoking gun they need for shutting down Humanities departments in universities everywhere, the company itself makes clear that using an algorithm to comprehend literature is a stepping stone to more practical uses. "For a computer to understand humans speaking in natural language and respond appropriately, it needs to capture and represent a large amount of knowledge that is not just words, but also common sense and context about the topic being discussed by the human," said Maluuba cofounder & CEO Sam Pasupalak. "Maluuba is working with leading experts and the world's premiere academic center for deep learning to design systems that can represent knowledge and answer questions in natural language.

Today we posted a new Tech Note in which Greg Diamos, a research scientist at Baidu's Silicon Valley AI Lab, discusses a new technique for speeding up the training of deep recurrent neural networks. At SVAIL, our mission is to create AI technology that lets us have a significant impact on hundreds of millions of people. We believe that a good way to do this is to improve the accuracy of speech recognition by scaling up deep learning algorithms on larger datasets than what has been done in the past. These algorithms are very compute intensive, so much so that the memory capacity and computational throughput of our systems limits the amount of data and the size of the neural network that we can train. So a big challenge is figuring out how to run deep learning algorithms more efficiently.

In a paper published in Nature on 28th January 2016, we describe a new approach to computer Go. This is the first time ever that a computer program "AlphaGo" has defeated a human professional player. The game of Go is widely viewed as an unsolved "grand challenge" for artificial intelligence. Games are a great testing ground for inventing smarter, more flexible algorithms that have the ability to tackle problems in ways similar to humans. The first classic game mastered by a computer was noughts and crosses (also known as tic-tac-toe) in 1952. But until now, one game has thwarted A.I. researchers: the ancient game of Go.