Ongeveer 10 uur geleden - Universiteit van Amsterdam (UvA) - Amsterdam
De Afdeling Algemene Rechtsleer zoekt een Universitair Hoofddocent Rechtstheorie/ Inleiding tot het Recht. Als Universitair Hoofddocent zul je …
Neuromorphic Computing through Sophisticated Photonics Integrated MatricesThis PhD position is opened within the Electro-Optical Communication Systems (ECO) group, department of Electrical Engineering and it is intended to explore the use of photonic integrated circuits for ...
Neuromorphic Computing through Sophisticated Photonics Integrated Matrices
This PhD position is opened within the Electro-Optical Communication Systems (ECO) group, department of Electrical Engineering and it is intended to explore the use of photonic integrated circuits for Neuromorphic Computing within the Zwaartekracht project on Integrated Nanophotonics.
The brain, unlike the von Neumann processors found in conventional computers, is very power efficient, extremely effective at certain computing tasks, and highly adaptable to novel simulations and environments. Artificial Neural Network is an algorithm that mimic the signal processing architecture in the brain (fig. 1a), and has recently received an explosion of interest. These methods have dramatically improved speech recognition, visual object recognition, object detection and genomics. Artificial networks can contain up to millions of units in each hidden layer (fig. 1b), which though make forward propagation the rate limiting step in many applications. Many efforts have been made to increase their computing speed, as is the case of neuromorphic computing.
Photonic platforms offer an alternative approach to microelectronics , being potentially able to outperform as for computing speed and the power efficiency. Indium phosphate based Integrated Photonics offers a reach platform in terms of amplification, non-linear effects and scalability, which are the main process features of an optical neural network. Based on the renowned expertise of TU/e ECO group on the use of sophisticated photonic integrated circuits (PIC) for telecom applications, we now would like to explore photonic integration for improving computation performance which can find application in heavy modeling for weather forecast and microfluidics, in fast reliable medical diagnostics as well as in telecoms.
The student has to research photonic integration solutions and architectures for neural networks and extract optical operational parallelism, energy consumption per operation and number of operation per seconds. The use of a networks based on matrices of photonic integrated devices can be used to implement matrices multiplications and exploit non-linearity. The photonic integration based neural network will be simulated and designed starting from the single photonic integrated neuron and moving to the full architecture, to tackle the scalability issue. Both analog and digital approaches will be considered for the best performance. The student will develop these new concepts also with the help of parallel experiments with in-house fabricated chips based on networks of optical switches for understanding their potential. Based on what is possible nowadays and on the simulation results, he will design new PIC systems for high performance computing. The realization of a final prototype for application demonstration in fast signal processing/image processing is foreseen in collaboration with other groups at TU/e.
Due to the high multidisciplinarity involved in this project, we wish to get in contact with candidates that like challenges and want to proof themselves in a four year project, which requires strong background primarily on photonic integration, but also skills in programming, algorithms, micro-electronics, chip characterization as well as creativity and independency. A strong physical, engineering and photonics background is strongly required. Candidates for this PhD position should have a Master's degree in Electrical Engineering, or similar (photonics, telecommunications, physics) with excellent grades.
- Strong background in photonic integration and their non-linearities
- Affinity with electronic computing
- Strong programming skills
- Creativity and originality