Deep_in_France

In recent years we have witnessed an explosion of successful applications of deep learning including speech recognition, automatic translation, self-driving cars, computers that can beat professional Go players, and recommender systems. Deep networks designed for these tasks have millions and billions of parameters that take enormous resources to train in terms of data, memory and computing power. Deep networks are clearly data, computationally and memory intensive, making them difficult to use and to deploy in particular on embedded systems. Applications in embedded devices ---such as smart phones, tablets, but also intelligent vehicles such as self-driving cars and drones, as well as wearable devices--- require low-power and memory efficient solutions to to solve recognition and scene understanding problems. Such requirements are not well aligned with the current resource-heavy approach to deep learning. Reducing the number of parameters and computational requirements, while preserving predictive performance, is critically important for deploying deep networks in this context. Deep in France collaborative research program aims at expanding the frontier of green deep learning. Green deep learning refers to the practice of using deep learning more efficiently while maintaining or increasing overall performance. Our vision is to develop theory and new deep learning architecture, algorithms and implementation allowing to deal with limited resources in terms of training examples and computing power. It will facilitate widespread use of deep learning to yet under-explored application domains such as audio scene recognition, embedded perception and video prediction. The bottlenecks regarding deep learning addressed by \projectname are both theoretical and methodological, as well as technical related to the optimization procedure and implementation. Our project also aims at bringing together complementary machine learning, computer vision and machine listening research groups working on deep learning with GPU’s2 in order to provide the community with the knowledge, the visibility and the tools that brings France among the key players in deep learning. The long-term vision of Deep in France is to open new frontiers and foster research towards algorithms capable of discovering sense in an automatic manner, a stepping stone before the more ambitious far-end goal of machine reasoning.