New lip-reading technology developed at the University of East Anglia (UEA) could help in solving crimes and provide communication assistance for people with hearing and speech impairments. The visual speech recognition technology, created by Dr Helen L. Bear and Prof Richard Harvey of UEA's School of Computing Sciences, can be applied "any place where the audio isn't good enough to determine what people are saying," Dr Bear said. Dr Bear, whose findings will be presented at the International Conference on Acoustics, Speech and Signal Processing (ICASSP) in Shanghai on March 25, said unique problems with determining speech arise when sound isn't available - such as on CCTV footage - or if the audio is inadequate and there aren't clues to give the context of a conversation. The sounds '/p/,' '/b/,' and '/m/' all look similar on the lips, but now the machine lip-reading classification technology can differentiate between the sounds for a more accurate translation. Dr Bear said: "We are still learning the science of visual speech and what it is people need to know to create a fool-proof recognition model for lip-reading, but this classification system improves upon previous lip-reading methods by using a novel training method for the classifiers. "Potentially, a robust lip-reading system could be applied in a number of situations, from criminal investigations to entertainment.

We discuss a general method to learn data representations from multiple tasks. We provide a justification for this method in both settings of multitask learning and learning-to-learn. The method is illustrated in detail in the special case of linear feature learning. Conditions on the theoretical advantage offered by multitask representation learning over independent task learning are established. In particular, focusing on the important example of half-space learning, we derive the regime in which multitask representation learning is beneficial over independent task learning, as a function of the sample size, the number of tasks and the intrinsic data dimensionality. Other potential applications of our results include multitask feature learning in reproducing kernel Hilbert spaces and multilayer, deep networks.

I've just finished Week 5 of the Coursera/Stanford Machine Learning course. It has been a mixture of refreshing, relearning, and new for me. I had already been using, building, and researching/evaluating machine learning algorithms for a number of years. I therefore felt like I'knew' a lot of the concepts, particularly the introductory ones. I put'knew' in quotes, however, since I've always had a feeling that I don't know them well enough, no matter how many times I've used them.

Linear regression is one of the simplest machine learning techniques you can use. It is often useful as a baseline relative to more powerful techniques. Like all regressions, we wish to map some input X to some input Y. You may recall from your high school studies that this is just the equation for a straight line. When X is 1-D, or when "Y has one explanatory variable", we call this "simple linear regression".

After a first all-day introduction to Deep Learning back in February, the Deep Learning class will continue in April and May. These classes will mostly be hands-on workshops (please don't forget to bring your laptop), with a minimum of formal theory to support these. Programming level: Beginner to advanced. Important: We will check IDs at the entrance. You will not be able to enter the school if you are not on the list. Please give us your real name when you register to this meetup.

In an announcement made on Wednesday at its GCP Next conference in San Francisco, tech giant Google said that it is rolling out the private beta of a new Cloud Machine Learning service which will enable businesses to create a custom machine learning model for predicting the future of their ventures. According to the details shared by Google, the Cloud Machine Learning service has the capability to handle data ingestion and training, and subsequently make use of the resultant machine-learning model to make predictions for a business' future. Google said that for building a custom machine learning model that can make future predictions for the future of a business, users of the Cloud Machine Learning service need to work with data which they have stored in Google's other cloud services. In a demonstration of the creation of a custom machine learning model for predicting the future of a business, Jeff Dean -- the chief of Google's Brain deep-learning research project -- showed how the Cloud Machine Learning service could build a model which predicts a click by a consumer on an advertisement. The model demonstrated by Dean, to show how the Cloud Machine Learning service works, was based on marketing software firm Criteo's anonymized data pertaining to consumers' chances of clicking on an advertisement.

Changing world: "As the first generation of self-driving cars and battlefield warbots filter into society, scientists are working to develop robots with moral decision-making skills." THE World Science Festival held in Brisbane in early March confirmed that robots, artificial intelligence and machine learning were now part of our lives. Thousands attending the festival came to watch, touch and play with cute, shiny robots capable of dodging objects, following commands and engaging in smart banter. However, if the future has arrived, now we have to deal with it. The World Science Festival was also an important forum as world experts discussed robot morality and ethics and what role we wanted robots to play in the future.

Marvin Minsky, an American scientist working in the field of artificial intelligence (AI) who co-founded vthe Massachusetts Institute of Technology (MIT) AI laboratory, wrote several books on AI and philosophy, and was honored with the ACM A.M. Turing Award, passed away on Sunday, Jan. 24, 2016 at the age of 88. Born in New York City, Minsky attended the Ethical Culture Fieldston School, the Bronx High School of Science, and Phillips Academy, before entering the U.S. Navy in 1944. After leaving the service, he attended Harvard University, where he earned a bachelor's degree in mathematics in 1950. He then went to Princeton University, where he built the first randomly wired neural network learning machine, the Stochastic Neural Analog Reinforcement Calculator (SNARC), before earning his Ph.D in mathematics there in 1954. Doctorate in hand, Minsky was admitted to the group of Junior Fellows at Harvard, where he invented the confocal scanning microscope for thick, light-scattering specimens, decades in advance of the lasers and computer power needed to make it useful; today, it is in wide use in the biological sciences.

U.S. President Barack Obama discussing his Computer Science for All plan to give students across the country the chance to learn computer science in school. In late January, U.S. President Barack Obama asked Congress to approve 4.1 billion in spending in the coming fiscal year to support the Computer Science for All initiative, aimed at providing computer science education in U.S. public schools. Obama pointed out computer science is no longer "an optional skill" in the modern economy," yet "only about a quarter of our K–12 (kindergarten through 12th grade) schools offer computer science. Twenty-two states don't even allow it to count toward a diploma." While many organizations have contributed to the national effort to see real computer science exist and count toward graduation requirements in U.S. public schools, former ACM CEO John R. White said, "ACM has been there from the beginning." Indeed, White contends Obama's Computer Science for All initiative "in a way represents the ...

Following in the wake of the recent trouncing of humans by artificial intelligence platform AlphaGo, Google has announced the launch of a cloud-based machine learning platform. The search giant's new large-scale platform will be able to learn and make predictions'across a whole variety of scenarios', and is reminiscent of the fictional Skynet service from Terminator. A limited preview of the service is now available for users to build their own machine-learning models'that work on any type of data, of any size'. Following in the wake of the recent trouncing of humans by artificial intelligence platform AlphaGo, Google has announced the launch of a cloud-based machine learning platform. The search giant's new large-scale platform will be able to learn and make predictions'across a whole variety of scenarios' A limited preview of the service is now available for users to build their own machine-learning models'that work on any type of data, of any size'.