Novel math could bring machine learning to the next level

EurekAlert  September 2, 2019 Before the neural network can begin to perform facial recognition, it is typically necessary to present it with thousands of faces. Much of these machines have been increasingly successful at pattern recognition but what goes on inside them as they learn their task is unknown. An international team of researchers (Portugal, Italy) has shown that artificial vision machines can learn to recognize complex images faster by using topological data analysis which was developed 25 years ago. Current neural networks are not good at topology. The team mathematically describe how to enforce certain symmetries, and this provides […]

Spotting objects amid clutter

MIT News   June 19, 2019 State-of-the-art algorithms can sift the bad associations from the good once features have been matched, but it is very slow. With a technique developed by researchers at MIT a robot can see the object through all this clutter and accurately pick out an object, such as a small animal, that is otherwise obscured within a dense cloud of dots, within seconds of receiving the visual data. Their technique prunes away outliers in polynomial time even for increasingly dense clouds of dots. The technique can thus quickly and accurately identify objects hidden in cluttered scenes. They […]

New program picks out targets in a crowd quickly and efficiently

Phys. org  February 22, 2019 Previous work on visual search has focused on searching for perfect matches of a target after extensive category-specific training. An international team of researchers (Singapore, USA – University of Minnesota) shows that humans can efficiently and invariantly search for natural objects in complex scenes. They developed a biologically inspired computational model that can locate targets without exhaustive sampling and generalize to novel objects. They trained the model to look for something that had similar features to the example image of a dog. This enabled the model to generalize from a single dog image, to the […]

Deep-learning technique reveals ‘invisible’ objects in the dark

Science Daily  December 12, 2018 Using deep neural network technique researchers at MIT reconstructed transparent objects from images of those objects, taken in almost pitch-black conditions. A computer was trained to recognize more than 10,000 transparent glass-like etchings, based on extremely grainy images of those patterns, with about one photon per pixel. They found that the computer learned to reconstruct the transparent object from the new grainy image, not included in the training data. The technique is of practical importance for medical imaging to lower the exposure of the patient to harmful radiation, and for astronomical imaging…read more. TECHNICAL ARTICLE