Transforming AI with the Power of Generative Networks
Dr. Emile-Reyn Engelbrecht and Professor Johan du Preez showed that generative networks used for semi-supervised learning and for recognising unseen categories work in the same way.
Materials, structures, energy, robotics and computing.
Dr. Emile-Reyn Engelbrecht and Professor Johan du Preez showed that generative networks used for semi-supervised learning and for recognising unseen categories work in the same way.
Professor Vladimir Kossov's team combined experiments and modelling to map when gas mixtures containing a greenhouse gas switch from slow diffusion to convective mixing.
Dr. Ilche Georgievski and Professor Marco Aiello present SH, an open-source AI planning system that breaks complex tasks into steps and connects to other software as a web service.
Dr. Simona Onori and Dr. William Chueh found that electric vehicle battery cells cycled with realistic, varying driving loads lasted substantially longer than cells tested under steady constant-current discharge.
Professor Yun Li's team built a grey-box model that combines physical laws with neural networks and found it modelled a cleanroom temperature control system better than conventional approaches.
Professor Becky Loo and Ting Lian used taxi GPS data to measure crash-caused traffic delays in Hong Kong, finding road networks and urban surroundings explain delays better than crash details.
Dr. Nabanita Saikia reviews how inorganic nanoparticles and biomaterials such as bioactive glass can serve as scaffolds helping stem cells rebuild bone, nerve, skin and cartilage, while noting open safety questions.
Professor Cai Wen's team built a radar network trained on simulated data and fine-tuned with a few real samples, which picked out weak moving targets amid clutter in simulations.
Gary Hahn and colleagues at Oak Ridge National Laboratory built Cyber Grid Guard, a ledger-secured backup system that detected and confirmed every tested fault type in a simulated substation.
Professor Ning Xi's team built an artificial muscle from gel-coated twisted strings that stiffens as it twists faster, letting it adjust stiffness and force in a way resembling human muscle.
Dr. Maryam Mobed-Miremadi's team designed printed hollow microneedles with a built-in stopper that limits how deep they penetrate, giving consistent depth in skin-like hydrogel tests and simulations.
Professor Bert Brouwers has built a turbulence model from general statistical principles rather than fitted constants, and its predictions for channel flow closely matched detailed direct numerical simulations.