Dog puppies understand humans in a way that their relatives like wolves just don't thanks to undergoing some 14,000 years of domestication, a study has concluded. Researchers from Duke University tested both dog and wolf puppies in a series of tests that involved locating hidden food by picking up on human clues. They found that dog puppies have similar social cognition abilities as human babies -- and are able, for example, to instinctively recognise pointing as communication. This superficially simple understanding is, in fact, rare in the animal kingdom. Not only did the wolf pups in the study lack it, but so do chimps, our closest relatives.

Two years ago, Gal Salomon's mother developed sepsis during a stay in the hospital. "It was a big hospital with a lot of patients and no one saw or understood it was happening," Salomon recalls bitterly. "We lost her after two days." So when Salomon, then a partner at Israeli venture capital firm Pitango, was introduced to Clew Medical, he knew immediately that he had to get involved. Clew develops software that uses artificial intelligence (AI) to predict which patients in a hospital's intensive care unit (ICU) are at the highest risk of imminent deterioration, and it alerts staff so they can intervene early.

Rarely does a day go by without more news predicting the end of work. After all, autonomous vehicles are all but certain to replace truckers and taxi drivers in the coming decades, and robots have already taken over many jobs in factories and warehouses, and will continue to expand their reach beyond heavy industry as they become smarter and ever more affordable. Perhaps most frighteningly, even professional services no longer seem safe from the encroachment of increasingly sophisticated artificial intelligence (AI). Law firms, for example, employ electronic-discovery software, which uses natural language processing to sift through reams of documents faster and more cheaply than the entry-level lawyers who used to do this tedious work. Deep-learning image recognition tools can flag and classify worrisome tumors in digital scans as well as, or better than, experienced radiologists.

A stability criterion for dynamic parameter adaptation is given. In the case of the learning rate of backpropagation, a class of stable algorithms is presented and studied, including a convergence proof.

A stability criterion for dynamic parameter adaptation is given. In the case of the learning rate of backpropagation, a class of stable algorithms is presented and studied, including a convergence proof.

A stability criterion for dynamic parameter adaptation is given. In the case of the learning rate of backpropagation, a class of stable algorithms is presented and studied, including a convergence proof. 1 INTRODUCTION All but a few learning algorithms employ one or more parameters that control the quality of learning. Backpropagation has its learning rate and momentum parameter; Boltzmannlearning uses a simulated annealing schedule; Kohonen learning a learning rate and a decay parameter; genetic algorithms probabilities, etc. The investigator always has to set the parameters to specific values when trying to solve a certain problem. Traditionally, the metaproblem of adjusting the parameters is solved by relying on a set of well-tested values of other problems or an intensive search for good parameter regions by restarting the experiment with different values. Inthis situation, a great deal of expertise and/or time for experiment design is required (as well as a huge amount of computing time).