This PhD position is within the Photonic Integration group of the Electrical Engineering Department and is part of a research line focusing on neuromorphic integrated photonics.
Information processing with artificial neural networks has re-emerged as the most important framework for data science and signal processing in recent years. Simulation of neural networks are performed with dedicated electronic hardware today which consumes a large amount of energy and are from an architectural point of view not well suited for neural computations. Photonics has been viewed as a promising hardware accelerator for certain types of neural computation, having the potential for huge speed improvements and energy savings. Semiconductor lasers show excitable temporal dynamics that equal that of biological spiking neurons and are able to simulate neuron behavior faster than electronics. Integration of such photonic neurons and scaling towards larger networks is thus very desirable but not yet well understood.
Research in the group has resulted in photonic circuit platforms which enable complex circuits and systems-on-chip for a wide range of applications. The work of the PhD researcher should use the technology as a vehicle to explore the integration of photonic neurons and its scalability. Lasers with optical feedback and injection have been demonstrated and studied before and their suitability to form photonic neurons is to be explored.
The following research questions shall be addressed:
- How can nonlinear laser dynamics be used to simulate photonic neuron behavior?
- What factors limit the scalability and cascadability of photonic neurons and how can we overcome those?