Ana Correia, Bruno Saraiva, Guilherme Carvalho, Hélder Avelar, Luís Outeiro and Pedro Machado (CTM)

Ana Correia, Bruno Saraiva, Guilherme Carvalho, Hélder Avelar, Luís Outeiro and Pedro Machado (CTM)

“The CTM coordinators would like to nominate the researchers Ana Correia, Bruno Saraiva, Guilherme Carvalho, Hélder Avelar, Luís Outeiro and Pedro Machado*. This nomination stems from the development of a CMOS integrated circuit for neuromorphic computing. The proposed architecture includes innovative aspects that are very relevant to the state of the art, particularly concerning the design of an efficient and precise time-amplitude conversion method. The team’s work took place within the scope of the THEIA project, in partnership with BOSCH, which represents a very relevant stage in the consolidation of research activities in advanced and highly efficient hardware for computing. This is an area of extreme importance at a global level, and it will contribute to boost the emergence of new Artificial Intelligence paradigms – which demonstrates the great relevance of the work developed by these researchers”.

– CTM coordinators 

This nomination stems from the development of a CMOS integrated circuit for neuromorphic computing. What does this circuit consist of?

It’s a neuromorphic circuit that, as the name suggests, aims to mimic the way the brain processes and transmits data. The circuit includes structures that act as the neurons and synapses we find in the human brain. The synapses were arranged in a crossbar structure and feature as inputs neuron impulses coming from a previous layer. As in a traditional neural network, inputs to posterior layer neurons correspond to a weighted sum of impulses, with said weight defined by synapses.

The architecture developed also presents a series of innovative aspects concerning the state of the art. What are said innovative aspects and how can they contribute to the emergence of new Artificial Intelligence paradigms?

Nowadays, there are many groups developing research in neuromorphic computing; however, we contribute with unique hardware, with weighing carried out in the temporal domain and with non-volatile memories, still maintaining a low energy footprint. The most innovative aspect of this circuit is the use of impulses with different widths to replace the traditional method of weighing in amplitudes. In this sense, the circuit doesn’t need to regularly resort to traditional digital-analogue converters, which represent higher area costs when compared to the other design elements. Another crucial aspect to the implementation is the location of memories within the processing units, speeding up data transfer between the parties.

Which aspects of your job do you enjoy the most?

Integrated circuits design is one of our passions; what motivates us the most is the opportunity to develop this work and reach a complex final product, and the sense of accomplishment. Another great aspect is the team spirit and solidarity that projects like this one tend to involve. We’re pretty sure that, after this project, we are a stronger and more united team.

How do you comment on this nomination?

It was a big surprise, and we are obviously very grateful for the acknowledgement. We would also like to thank Prof. Vítor Grade Tavares and Prof. Manuel Cândido Duarte, the mentors of this project – who dedicated much of their time to ensure the development of this integrated circuit.

* the researcher Pedro Machado is no longer associated with INESC TEC.