Master’s student

Fabrication and Characterization of Oscillating Neural Networks

Ref. 2021_058

Context

The desire to mimic brain information processing has increasingly motivated research efforts towards Artificial Intelligence and Neuromorphic Computing [1]. Most commonly used AI-powered applications are digital implementations of Artificial Neural Networks (ANNs). The digital approach is however orders of magnitude less energy-efficient than the human brain and offers less performant learning procedures due to a lack of in-circuit programmability –i.e. plasticity [2].

Hardware accelerators based on reprogrammable synaptic weights, Spiking Neural Networks (SNNs) or even oscillators seem to offer new solutions by emulating physically implemented Recurrent and Convolutional Neural Networks (CNN) [3]; already popular for digital image classification tasks. The latter can act as fundamental blocks in a network, where compact oscillating devices can be realized by exploiting the insulator-to-metal transition of VO2[4]. Our Materials Integration and Nanoscale Devices (MIND) group at IBM Research Europe in Zurich is currently investigating the computational potential of this technology within the framework of the NeurONN project and would like you to join for a minimum of 6 months to contribute to the experimental aspect of this work.

 

Description

Our objective is to demonstrate pattern recognition by designing analog oscillator circuits with associative memory capabilities [4]. In this project, we wish to explore novel device concepts through thermal and electrical characterization of our VO2-based oscillators, and evaluate improved fabrication techniques to integrate our design into a power-efficient smart circuit. You will have access to world-class research infrastructures, be offered meaningful experience in the ETH/IBM BRNC cleanroom, and have the opportunity to gain knowledge on instrument measurements followed by data analysis. You will join a dynamic and international team where new ideas and cooperation are valued.

 

Requirements

  • Pursuing a Master’s degree in Electrical Engineering, Computer Science, Neuroscience, or a related field
  • Enrolled in a Swiss or EU institution – for Student Visa purposes
  • Proficiency in English, both written and oral
  • Programming skills in MATLAB and/or Python
  • Experience in Material Science is a plus
  • Background in Machine/Deep Learning or Neural Networks theory desired
  • Prior internship or industrial experience preferred

 

Diversity

IBM is committed to diversity at the workplace. With us you will find an open, multicultural environment. Excellent flexible working arrangements enable all genders to strike the desired balance between their professional development and their personal lives.

 

COVID risk management

The access to the Zurich lab and BRNC cleanroom is currently not restricted. In case of a lockdown, VPN access allows for remote use of many of the services provided on-site with software licenses or cloud computing. Some experiments can be performed remotely. Otherwise, simulations represent another valuable research direction with relevant and interesting content in the scope of a Master’s project.

 

Please note that this is a Master's Thesis project, not an internship, and will not be compensated financially.

 

How to apply

If you are interested in this role, please provide a Resume, a Motivation Letter, and your Bachelor’s degree as well as your most recent transcripts.

 

For more information on technical questions please contact
Dr. Siegfried Karg () or Olivier Maher ().

 

References

[1] D. Heger and K. Krischer, "Robust autoassociative memory with coupled networks of Kuramoto-type oscillators," Physical Review, 2016
[2] G. Indiveri and S.-C. Liu, "Memory and Information Processing in Neuromorphic Systems," Proceedings of the IEEE, vol. 103, no. 8, pp. 1379 -1397, 2015.
[3] E. Chicca, F. Stefanini, C. Bartolozzi and G. Indiveri, "Neuromorphic Electronic Circuits for Building Autonomous Cognitive Systems," Proceedings of the IEEE, vol. 12, no. 9, pp. 1367 -1388, 2014.
[4] E. Corti and S. Karg, "Coupled VO2 oscillators circuit as analog first layer filter in convolutional neural networks," Frontiers in Neuroscience, 11 February 2021.

Illustration

Illustration