Nature

Cross-layer transmission realized by light-emitting memristor for constructing ultra-deep neural network with transfer learning ability

Deep neural networks (DNNs) possess the capability to represent more complex nonlinear problems than shallow neural networks, and their distributed data learning method is more effective 1,2,3. The ...
Nature

Supernetwork-based efficient mapping of deep learning applications to mixed-precision hardware using model adaptation

The rapid proliferation of Artificial Intelligence applications necessitates scalable solutions that perform efficiently under real-world constraints. Heterogeneous accelerators combining specialized ...
Science Daily

Training algorithm breaks barriers to deep physical neural networks

Researchers have developed an algorithm to train an analog neural network just as accurately as a digital one, enabling the development of more efficient alternatives to power-hungry deep learning ...