The Optical Networking and Communication
Conference & Exhibition

San Diego Convention Center,
San Diego, California, USA

Optics for Neuromorphic Computing and Machine Learning: Status, Prospects and Challenges

Organizer:

Paraskevas Bakopoulos, Mellanox Technologies, Greece; Shastri Bhavin, Queen's University, Canada; Chigo Okonkwo, TU Eindhoven, Netherlands
 

Description:

The recent rise in artificial intelligence and neuromorphic computing has demonstrated super-human performance in tasks such as image recognition, language translation, cancer detection, healthcare, self-driving cars, etc. This rise can be attributed to algorithmic innovations, access to big data, and new hardware (GPUs, Google’s tensor processing unit). With more computing applications, new demands are being placed on hardware that are faster and energy efficient. Recently, there has been a resurging interest in using light to build processors to meet these demands and potentially enable new applications in high-performance computing, solving optimization problems, accelerating deep learning, etc. Photonic technologies offer high-speed optical communication and massive parallelism with optical signals, coupled with the advances in photonic integration technology and a large-scale silicon industrial ecosystem. 

Through a collection of talks and panel discussions, this workshop will cover topics on the current status of the field in using light for machine learning and neuromorphic computing. The applications domains that may drive the demand for photonic and optoelectronic solutions, and the challenges associated with commercializing this technology will be addressed. The topics will range from devices, systems, architectures, algorithms, and applications for: photonic reservoir computing with delay-based system; multiwavelength and coherent optical neural networks with integrated photonics; optical spiking neural networks with excitable lasers and phase-change materials; free-space diffractive optics; and coherent Ising machines.

Sponsored by: