Conferences and supporting programme
Implementing Artificial Neural Networks, Unleashing New Possibilities of Edge Intelligence
The rapidly-growing area of artificial intelligence (AI), neural networks (NN) and machine learning offers tremendous promise as developers attempt to bring higher levels of intelligence to their systems. NNs are a paradigm that engineers have been using to implement systems that can continuously learn and infer, based on the learning. Computational requirements for such a system vary widely depending upon application. Traditionally deep learning techniques are used in the data center and typically rely on large, high performance GPUs to meet those demanding computational requirements. Designers extending the advantages of AI to the network edge don’t have the luxury of power-hungry GPUs. At the edge, the deep learning techniques that use floating-point computation in the data center are impractical. Designers must develop computationally efficient solutions that not only meet accuracy targets, but also comply with the power, size and cost constraints of the consumer market. This session will highlight “on-device Artificial Intelligence (AI),” which uses the NN models to compare new incoming data and infer intelligence. On-device AI provides improved user privacy, since mass amounts of personal data is not sent to the cloud as processing occurs locally, on-premise. We will evaluate technologies such as FPGAs that make edge computing possible and are needed to maximize parallel computing, as well as explain the intelligence that these low power solutions bring to battery powered applications. We will also discuss how building AI into an FPGA running open source RISC-V processor with some accelerators does not only cut power consumption, but also accelerate response time in addition to keeping processing local, improving security and privacy. By the end of this session, designers will learn how to bring the advantages of AI, neural networks and machine learning to resource-constrained, power-optimized network edge devices.
--- Date: 27.02.2018 Time: 9:30 AM - 10:00 AM Location: Conference Counter NCC Ost