Today, many problems encountered in the fields of image processing and signal processing are tackled by using deep artificial neural networks due to their high performance. This requires training a wide range of artificial neural network models in problem specific data sets. The performance that can be obtained as a result of the training depends on the chosen artificial neural network model as well as the training method.
Diversis has chosen deep learning optimization as its core technology and actively conducts research and developments efforts to improve the effectiveness of deep learning training techniques. The first tangible result of these efforts is our Deep Optimizer Framework.