Inspired by the fast processing of information in the human brain, a consortium of European research institutions is developing a novel kind of computing. The EU funded project PHOCUS that started in January aims to design photonic systems, communicating via light, to quickly perform complex computations including the rapid processing of large amounts of data, potentially consuming far less power than current supercomputers.
Recognizing a friend's face in a crowd of people is a complex task for the brain, yet it might take only a fraction of a second. "It is one of the biggest unresolved questions in brain research, how the electrical discharges of billions of neurons are organised to deliver correct answers in such a short time," says Claudio Mirasso, the project coordinator of the Universitat de les Illes Balears. "However, in recent years a paradigm has been developed in neuroscience that might lead to an answer," adds Mirasso.
Neuroscientists exploit an analogy between the human brain's response to external stimuli and the reaction of a liquid to external perturbations, such as a pebble thrown into water. From the waves generated by the impact, one can conclude where and when the pebble hit the surface. Similarly, it could be possible to draw information about external stimuli from the transient responses of neural networks. These networks into which stimuli or inputs are fed are referred to as "reservoirs."
However, the existing computer models of neural networks are difficult to train since the interactions among the model's elements have to be carefully re-adjusted for different inputs. The concept of reservoir computing avoids this problem by leaving the reservoir unaffected and training only the readout of processed data. "Preliminary experiments indicate that this is far easier than training the reservoir itself," says Jurgen Kurths of the Potsdam-Institute for Climate Impact Research.
So how does reservoir computing work? Like a pebble rippling the water surface, external stimuli or inputs remain detectable in the reservoir for a certain time. This memory of the input, in combination with the emerging response of the reservoir, transforms the input into a large number of dynamical states of the reservoir, thus creating a high-dimensional state space. The trick of reservoir computing is that the reservoir's responses to different inputs are easier to identify in this high-dimensional state space than in the original, lower dimensional input space " the number of dimensions of the input space corresponding to the number of features necessary to recognise a known face, for example. It has been shown that the identification in the high-dimensional state space can be used to classify different inputs.
Complex collective behaviour exhibited by coupled nonlinear dynamical systems — in this case, the reservoir's memory of the input and its emerging response — is at the core of this concept. The understanding of coupled complex systems' dynamics and in particular their synchronization properties has been significantly advanced in recent years. Photonic systems played a key role in these studies and demonstrate the usefulness of complex behaviour. Scientific interactions between laser and nonlinear dynamics physicists, neurophysiologists and brain researchers, led to the idea that photonic systems could be used to understand and eventually mimic some of the functionalities of the brain.
The realization of reservoir computing using photonic systems offers great opportunities but imposes even more challenges. Photonic systems allow for extremely fast processing and are compatible with telecommunication technology. Integration of larger photonic systems is, however, technologically challenging and expensive.
The PHOCUS consortium has identified a novel approach to realize the functionality of a complex network with only few photonic components. The basic idea is to utilize the generically occurring delays in feedback and coupling of one or a few semiconductor lasers to generate a very high-dimensional state space. Ultimately, PHOCUS aims at photonic implementations of reservoir computing operating at high data rates. This would offer alternatives to supercomputers and computer clusters for specific tasks requiring reduced size and less power consumption.