Prof. Giacomo Indiveri

 

Main Goals, Keywords

We are investigating the computational properties of spike-based neural circuits using theoretical models, software simulation tools, and neuromorphic VLSI circuits and systems. In particular we are interested in soft Winner-Take-All networks, cooperative-competitive networks, spike-based plasticity mechanisms, selective attention mechanisms, vision sensors, and their implementation in VLSI technology.

Group Members

1 Professor, 2 Post-doctoral fellows, 11 PhD students

Previous and Current Research

The neuromorphic approach implements the principles of computation used by the nervous systems in analog/digital VLSI technology, exploiting the device physics of transistors to reproduce the biophysics of neural cells.
Neuromorphic VLSI devices model the properties of biological systems down to the single neuron and synapse dynamics and use a digital asynchronous communication infrastructure to transmit spikes across chips and to computers and robots.
We are currently investigating spike-based learning algorithms and VLSI circuits for real-time classification of sensory data (such as speech, or visual stimuli), implementing selective-attention architectures and vision sensors for active vision systems, and investigating auditory scene analysis using neuromorphic multi-chip systems.

Future Projects

We plan to develop large-scale, distributed, multi-chip VLSI networks of biologically realistic neural networks, with adaptation and learning.
These multi-chip systems will be used to explore the computational properties of spiking neural circuits, and to implement real-time behaving systems for practical applications.

Techniques and Equipment

For analyzing and simulating neural networks we use standard SW tools, ranging from Matlab to Python.
For designing VLSI chips we use the set of Tanner Tools provided via the Europractice Service. VLSI chips are fabricated via both Europractice and MOSIS services. Typically, the VLSI chips are implemented using standard 0.35u CMOS technology, and occupy areas of 10 to 20 square millimeters.

Selected Publications

• Kasabov N, Dhoble K, Nuntalid N, Indiveri G. Dynamic evolving spiking neural networks for on-line spatio- and spectro-temporal pattern recognition. Neural Netw. 2013 May;41:188-201. doi: 10.1016/j.neunet.2012.11.014. Epub 2012 Dec 20.
• Neftci E, Toth B, Indiveri G, Abarbane H. Dynamic State and Parameter Estimation applied to Neuromorphic Systems. Neural Computation, 24:(7) 1669-1694, 2012
• Sheik S, Chicca E, Indiveri, G. Exploiting device mismatch in neuromorphic VLSI systems to implement axonal delays, International Joint Conference on Neural Networks (IJCNN) 1-6, 2012
• Sheik S, Coath M, Indiveri G, Denham SL, Wennekers T, Chicca E. Emergent Auditory Feature Tuning in a Real-Time Neuromorphic VLSI System. Front Neurosci. 2012;6:17. Epub 2012 Feb 6.

Selected Lectures, Seminars or Colloquia

The Telluride Neuromorphic Cognition Engineering Workshop (http://ine-web.org)
The Capo Caccia Workshops toward Cognitive Neuromorphic Engineering (http://cne.ini.uzh.ch)

Funding

EU ERC, Swiss National Science Foundation, EU FP7

URL

http://ncs.ethz.ch/