[PhD] Visual perception by neural fields on a neuromorphic chip

Supervision

BISCUIT team, Loria
Supervisor : Bernard Girau (HDR)
Bernard.Girau@loria.fr

Description

Scientific context

Among various alternatives to standard Von Neumann architectures, neuromorphic approaches currently benefit from the recent application successes of deep learning methods and the implication of major semiconductor companies through impressive neuromorphic chips such as IBM TrueNorth and the research prototype chip from Intel called Loihi (https://www.intel.com/content/www/us/en/research/neuromorphic-computing.html). With a Loihi-based system, up to 8 billions of spiking neurons with on-chip learning can be implemented in a fully parallel way. We have initiated an INRC (Intel Neuromorphic Research Community)  project in which we intend to adapt our neural field models for visual perception to the innovative Intel Loihi chip architecture.

The subject proposed here is linked to the definition and hardware simulation of spiking dynamic neural fields (2D maps of integrate-and-fire neurons) on a neuromorphic chip. Dynamic neural fields (DNF) model the continuous interaction between excitatory and inhibitory neurons in a cortical population. DNF have been successfully applied to visual memory, tracking, selection, scene exploration, human motion discrimination, etc. Because of the dense interconnection scheme of DNF, their computational requirements in terms of bandwidth and co-localized computation-memory needs, make them difficult to implement in real-world scenarios on conventional computers when real-time or low-consumption constraints exist. It is thus tempting to implement them in specialized hardware.

Detailed description

The goal of this thesis is to adapt neural field models to digital neuromorphic chips, and more especially to the Loihi architecture. Different aspects must be dealt with.

1- Definition and analysis of a DNF model where each neural unit uses the kind of spiking computation that can be implemented in the Loihi neuromorphic cores. In the field of vision, we have already shown that spiking DNF are able to select and track centers of interest in visual scenes, despite the presence of a strong level of noise or distractors (cf [1]), and we have derived an FPGA implementation coupled with an event-based DVS camera in [2].

2- Communication of neurons within neuromorphic cores and between them, knowing that the usually fully connected interconnection scheme of DNF makes it unable to be mapped in a distributed way onto a fundamentally 2D hardware substrate or rapidly induces communication bottlenecks in AER-based protocols when the size of the DNF increases. We have already explored the combination of DNF with cellular computing properties that can offer a more scalable solution, first showing that a randomized propagation of spikes was able to simulate the same kind of synaptic weights as in usual DNF even when neurons are only connected to their 4 immediate neighbors [3], then mixing this spiking and randomized approach with the bit-stream arithmetic principles to define compact spike-stream operators [4]. Yet such approaches do not directly fit neuromorphic chips.

3- Connection of multiple DNF to implement more complex neural field architectures, such as used for visual memory. This kind of connection is usually based on the notion of gaussian receptive fields between afferent DNF. An efficient way to map this kind of afferent connection to a neural map must be found on the Loihi architecture.

4- The Loihi chip features synaptic weight learning based on STDP (spike-timing dependent plasticity). Though most DNF applications use predefined weights, we have developed multi-map DNF models that learn their lateral synaptic weights with respect to the visual task to perform. These models have not yet been adapted to spiking computations. The on-chip learning capability of the Loihi chip motivates the definition of efficient algorithms to learn the synaptic weights of spiking DNF by means of STDP and to map them on neuromorphic chips.

Skills required

The candidate should have the equivalent of a Master in Computer Science, preferably in a specialty related to artificial intelligence and/or distributed numerical computation. Adequate knowledge of digital hardware design will be assessed, as well as experience in software design. The candidate must fluently speak English and/or French.

Bibliography

[1] R. Vazquez, B. Girau, J.-C. Quinton. Visual attention using spiking neural maps. International Joint Conference on Neural Networks IJCNN 2011, Ali Minai, Hava Siegelmann, Jul 2011, San José, United States.

[2] B. Chappet de Vangel, C. Torres-Huitzil, B. Girau. Event based visual attention with dynamic neural field on FPGA. International Conference on Distributed Smart Camera, Sep 2016, Paris, France.

[3] B. Chappet de Vangel, C. Torres-Huitzil, B. Girau. Randomly spiking dynamic neural fields. Journal of Emerging Technologies in Computing Systems, ACM, 2014.

[4] B. Chappet de Vangel, C. Torres-Huitzil, B. Girau. Stochastic and Asynchronous Spiking Dynamic Neural Fields. International Joint Conference on Neural Networks (IJCNN 2015), Jul 2015, Killarney, Ireland.

Keywords

Neural networks, neural fields, neuromorphic chip, digital hardware

 

Logo d'Inria