This 3-hour course (video slides) offers developers a quick introduction to deep-learning fundamentals, with some TensorFlow thrown into the bargain. Deep learning (aka neural networks) is a popular approach to building machine-learning models that is capturing developer imagination. If you want to acquire deep-learning skills but lack the time, I feel your pain. In university, I had a math teacher who would yell at me, "Mr. Görner, integrals are taught in kindergarten!"

Founding Attorney Chad Van Horn says, "We are excited to add ROSS to our already busy team of bankruptcy attorneys. ROSS will allow us to continue to be at the forefront of assisting consumers with their legal issues while keeping attorneys fees at an affordable level. ROSS will assist in equaling the playing field between boutique law firms and big law. In short, we're excited to add ROSS to our extremely talented team of legal professionals." ROSS is an artificial intelligence, and as such has never been to law school or passed the bar, but his research is impeccable.

Technology in the so-called Smart Machine Age, which includes AI, virtual reality and robotics, will bring huge changes not just in headcount, but also in how people innovate and collaborate. That will require new approaches to how people think, listen and relate, says Edward D. Hess, a professor of business administration at the University of Virginia. In the "Smart Age" now evolving, ego has no place. Instead, the focus will need to be on the quality of ideas, accuracy, emotional intelligence and mindfulness. Hess writes about these issues in the just-released book he co-authored with Katherine Ludwig, titled Humility is the New Smart: Rethinking Human Excellence in the Smart Machine Age. Hess discussed his ideas on the Knowledge@Wharton show on Wharton Business Radio, SiriusXM channel 111.

With the advent of large labelled datasets and highcapacity models, the performance of machine vision systems has been improving rapidly. However, the technology has still major limitations, starting from the fact that different vision problems are still solved by different models, trained from scratch or fine-tuned on the target data. The human visual system, in stark contrast, learns a universal representation for vision in the early life of an individual. This representation works well for an enormous variety of vision problems, with little or no change, with the major advantage of requiring little training data to solve any of them. In this paper we investigate whether neural networks may work as universal representations by studying their capacity in relation to the "size" of a large combination of vision problems. We do so by showing that a single neural network can learn simultaneously several very different visual domains (from sketches to planktons and MNIST digits) as well as, or better than, a number of specialized networks. However, we also show that this requires to carefully normalize the information in the network, by using domainspecific scaling factors or, more generically, by using an instance normalization layer.

This paper presents privileged multi-label learning (PrML) to explore and exploit the relationship between labels in multi-label learning problems. We suggest that for each individual label, it cannot only be implicitly connected with other labels via the low-rank constraint over label predictors, but also its performance on examples can receive the explicit comments from other labels together acting as an \emph{Oracle teacher}. We generate privileged label feature for each example and its individual label, and then integrate it into the framework of low-rank based multi-label learning. The proposed algorithm can therefore comprehensively explore and exploit label relationships by inheriting all the merits of privileged information and low-rank constraints. We show that PrML can be efficiently solved by dual coordinate descent algorithm using iterative optimization strategy with cheap updates. Experiments on benchmark datasets show that through privileged label features, the performance can be significantly improved and PrML is superior to several competing methods in most cases.

The ability to learn tasks in a sequential fashion is crucial to the development of artificial intelligence. Neural networks are not, in general, capable of this and it has been widely thought that catastrophic forgetting is an inevitable feature of connectionist models. We show that it is possible to overcome this limitation and train networks that can maintain expertise on tasks which they have not experienced for a long time. Our approach remembers old tasks by selectively slowing down learning on the weights important for those tasks. We demonstrate our approach is scalable and effective by solving a set of classification tasks based on the MNIST hand written digit dataset and by learning several Atari 2600 games sequentially.

COMPUTERS can already hold a massive amount of instantly retrievable data in a manner that puts most humans to shame, but getting them to actually display intelligence is an entirely different challenge. Now a team of researchers from Northwestern University just made a huge stride toward that goal with a computational model that actually outperforms the average American adult in a standard intelligence test. As PhysOrg reports, the witty computer system utilizes an AI platform called CogSketch that gives it the power to solve visual problems just by looking at them, which is something that has traditionally held back many examples of artificial intelligence, reports the New York Post. Being able to visually understand, interpret, and then use that data to come to a solution brings the computer system closer to the functioning of the human brain than many before it, and so the team pitted its creation against a popular standardised test called Raven's Progressive Matrices. The Raven's test (or RPM for short) is composed of 60 multiple-choice questions that measure the taker's ability to reason, using visual puzzles.

The following script is from "Artificial Intelligence," which aired on Oct. 9, 2016. Charlie Rose is the correspondent. The search to improve and eventually perfect artificial intelligence is driving the research labs of some of the most advanced and best-known American corporations. They are investing billions of dollars and many of their best scientific minds in pursuit of that goal. All that money and manpower has begun to pay off. In the past few years, artificial intelligence -- or A.I. -- has taken a big leap -- making important strides in areas like medicine and military technology. What was once in the realm of science fiction has become day-to-day reality. You'll find A.I. routinely in your smart phone, in your car, in your household appliances and it is on the verge of changing everything. On 60 Minutes Overtime, Charlie Rose explores the labs at Carnegie Mellon on the cutting edge of A.I. See robots learning to go where humans can'... It was, for decades, primitive technology.

UC Berkeley researchers are tapping the cognitive smarts of babies, toddlers and preschoolers to program computers to think more like humans. "Children are the greatest learning machines in the universe. Imagine if computers could learn as much and as quickly as they do," said Alison Gopnik a developmental psychologist at UC Berkeley and author of "The Scientist in the Crib" and "The Philosophical Baby." In a wide range of experiments involving lollipops, flashing and spinning toys, and music makers, among other props, UC Berkeley researchers are finding that children -- at younger and younger ages -- are testing hypotheses, detecting statistical patterns and drawing conclusions while constantly adapting to changes. "Young children are capable of solving problems that still pose a challenge for computers, such as learning languages and figuring out causal relationships," said Tom Griffiths, director of UC Berkeley's Computational Cognitive Science Lab. "We are hoping to make computers smarter by making them a little more like children."

This post is authored by Kristin M. Tolle, Director of Program Management for Advanced Analytics Ecosystem Development and Training at Microsoft. Cortana Intelligence, Microsoft's end-to-end platform for Advanced Analytics, offers a suite of services to solve real world customer problems. The suite has many moving parts – Data Lake, HDInsight (Hadoop), Event Hub, Machine Learning and R – just to name a few, and we realize it may be challenging for some of you to experience first-hand how all these services work together in concert. My team, which is tasked with training our partners to use these services to address their customers' needs, is keenly aware of the breadth of that knowledge surface area. In this blog post, I outline some of the best ways for you to learn about all things Big Data and Advanced Analytics from Microsoft, including many hands-on training options, and also how to stay in the loop on our future offerings.