Heterotic computing: exploiting hybrid computational devices

Hybrid quantum computers: a "one for all and all for one" approach with superconductors and spin ensembles

Professor Klaus Mølmer, University of Aarhus, Denmark

Abstract

Since the 1994 discovery by Peter Shor that a quantum computer may factor large numbers efficiently, the potential for quantum computing has been recognized by a variety of public, strategic, and commercial organizations.

Quantum computing may be implemented with physical components that are already studied extensively in the laboratory: trapped ions, cold atoms, superconducting circuits, liquid and solid state spin ensembles, etc., and elementary gate operations and algorithms have already been demonstrated in experiments.

The basic challenge for quantum computing is to find and control a physical quantum system that offers rapid processing, long-time storage, scalability to a sufficient processor size, and means for intermediate or long distance communication. Since no single quantum system meets all these requirements, the concept of hybrid quantum technologies has emerged, where the different tasks are shared between physical components that are individually optimized for the different functions.

The interfacing of physical systems with very different spatial, temporal and energetic properties presents a big challenge in itself. By means of a recent successful example involving superconducting circuits and atomic spin degrees of freedom I shall give an illustrative example of how that challenge may be met.

NMR classical computation (expt)

Dr Matthias Bechmann, Johannes Kepler University Linz, Austria

Abstract

Nuclear magnetic resonance (NMR) spectroscopy has been very successful, both for its role as spectroscopic tool to determine molecular structure on one side and as a test case for the quantum mechanical description of spins and their dynamics on the other. Precise measurement of the dynamics of spin-system ensembles is today facilitated by a high level of hardware engineering that has been put into the hardware. This all together made NMR also very attractive as a means for researching quantum computing.

Here the potential of such NMR spectrometers to implement classical computational paradigms is demonstrated. This scenario then makes it possible to combine and assess quantum and classical contributions to "computation" in a single experimental set-up.

Quantum and classical resources in measurement-based quantum computation

Dr Janet Anders, University of Exeter, UK

Abstract

Quantum physics is known to allow the implementation of powerful algorithms [1], however, the precise power increase over classical computation remains open. I will describe an insightful example where the computation relies on the interplay of two components - a restricted classical computer and its interaction with quantum correlated information. It turns out that quantum correlations can enhance the classical computer in such a way that it is able to compute problems beyond its own capability [2]. The example belongs to a type of computation, measurement-based quantum computation, that has no counterpart in classical computing as the computation is driven by quantum measurements rather than gates. Future implementations of this model may require the combination of mobile information carriers that sequentially interact with a stationary register [3] or adiabatic implementations, where the random measurement is replaced by a deterministic evolution [4].

References:
[1]. Anders and Wiesner, Chaos 21, 037102 (2011)
[2]. Anders and Browne, Phys Rev Lett 103, 070502 (2009)
[3]. Kashefi et al, Theoretical Computer Science 430, 51 (2012)
[4]. Antonio et al, arxiv 1309.1443 (2013)

Rapporteur

Professor Tony Hey CBE FREng, Microsoft Research, USA

Theory and practice of molecular computing with DNA

Dr Damien Woods, Caltech, USA

Abstract

One of the main aims of molecular engineering and computation is to control the structure and dynamics of molecular systems at the nanoscale. We'd like to design systems that are robust to fluctuations in temperature, fluid flow and other uncontrolled factors, yet consist of millions of interacting components. DNA is a versatile and programmable material that meets these criteria. The kinetics and thermodynamics of DNA are reasonably well-understood, and through straightforward Watson-Crick base-pairing interactions we can program this material to create complicated shapes and patterns, as well as to have intricate, even algorithmic, chemical dynamics, all at nanoscale spatial resolution. In this talk, I will give an overview of the state of the art with an emphasis on programability, abilities and limitations of these chemical systems. I will describe how computer science and biology can inspire us about which molecular systems we should try to build next, and how we can use mathematical and algorithmic thinking to give us the tools to control a cacophony of interacting molecules by simply letting them interact in a hands-off self-assembling fashion.

Physarum computing: from reaction-diffusion to slime mould

Professor Andrew I Adamatzky, University of the West of England, UK

Abstract

Natural systems show numerous, often unconventional, ways of information processing, which are adopted, exploited, mimicked in computer software and engineering devices.  Despite the profound potential offered by unconventional computing, only a handful of experimental prototypes are reported so far, for example gas-discharge analog path finder; maze-solving micro-fluidic circuits; geometrically constrained chemical computers; chemical reaction--diffusion processors; maze-solving chemo-tactic droplets; enzyme-based logical circuits; spatially extended crystallization computers for optimization and computational geometry; molecular logical gates  and circuit. In my talk I will discuss three families of growing pattern based computing devices: reaction-diffusion computers, crystallisation-based computers and slime mould computers. I will demonstrate how classical tasks of computational geometry and optimisation (Voronoi diagram, Delaunay triangulation, spanning trees, relative neighbourhood graphs, beta-skeletons) can be approximated by excitation (Belousov-Zhabotinsky medium), crystallisation in supersaturated solutions and biological growth patterns in slime mould of P. polycephalum propagating in a quasi two-dimensional space and interacting with each other.  I will explain how  logical circuits can be implemented using collision-based computing paradigm executed in simulated and experimental laboratory non-linear media computers; and, present experimental designs of binary adders implemented in excitable chemical medium and slime mould.

Rapporteur

Professor Ottoline Leyser CBE FRS, Sainsbury Laboratory, University of Cambridge UK

Self-similarity and recursion in algorithmic DNA self-assembly

Professor Natasha Jonoska, University of South Florida, USA

Abstract

Recursive processes and self-similarity of form and function can be seen in many natural phenomena. Although controlled algorithmic self-assembly, in particular DNA self-assembly has shown grand progress in recent years, controlling the assembly to achieve recursive growth has been difficult. In this presentation we will discuss recent advances in bimolecular computing. Then we will show a model where the building blocks of the assembly process are capable of transmitting and receiving binding site activation signals thereby  accomplishing  assembly in stages. Within this model, we show how  a recursive assembly of archetypal self-similar aperiodic structures can be realized. We also present recent experimental results supporting the model.

Toward a 'Siliconeural' computer: technological successes and challenges

Professor Alan Murray, University of Edinburgh, UK

Abstract

“Neural Networks” were, in the 1980s, viewed, very naively, as a potential panacea for all computational problems that did not fit well with conventional computing.  The field bifurcated during the 1990s into a highly-successful and much more realistic machine learning community and an equally pragmatic biologically-oriented “Neuromorphic computing” community.  Algorithms found in nature that use the non-synchronous, spiking nature on neuronal signals have been found to be (a) implementable efficiently in silicon and (b) computationally useful.  As a result, interest has grown in techniques that could create mixed “siliconeural” computers.  This talk will describe the computational path that leads to this research focus, which also has clear implications for prosthetics and “bionics”.  The challenges and successes in both (neuron-on-silicon) cell guidance and cell interfacing will be described, along with the remaining difficult questions that this effort poses.

Chemical computing with Belousov-Zhabotinsky Reaction

Professor Jerzy Górecki, Institute of Physical Chemistry of the Polish Academy of Sciences Warsaw, Poland

Abstract

My presentation will be concerned with information processing based on a specific type of highly nonlinear chemical photosensitive medium. A photosensitive variant of Belousov-Zhabotinsky (BZ) reaction has been considered as an interesting substrate for unconventional computing since the publication of image processing algorithm proposed by Kuhnert, Agladze and Krinsky in 1989. In my opinion this substrate nicely fits into the field of heterotic computing because information processing operations are performed by a chemical reaction, whereas input, readout and feedback can be executed through system illumination. In the following years more information processing applications of photosensitive BZ reaction have been studied. The basic signal processing devices, like logic gates or signal filters, can be made with non-homogenously illuminated medium. The recent studies on lipid droplets containing reagents of BZ reaction bring another dimension to the field. In early experiments the geometrical structure of illuminated and non-illuminated regions were generated by an experimentalist. Droplets are mobile thus the required structure can spontaneously appear at a specific nonequilibrium state. During my presentation I will discuss the perspectives of such approach.

Evolution in materio: evolving computation in materials

Dr Julian Miller, University of York, UK

Abstract

Can digital computers help us create new analogue computational devices?

Evolution in materio is concerned with the manipulation of a physical system by computer controlled evolution (CCE). It is a general method for obtaining analogue computational devices by utilizing physical effects that a human programmer need not be aware of.

Although a form of this methodology was hinted at in some work of Gordon Pask in the late 1950s it was not convincingly demonstrated until 1996 by Adrian Thompson, who showed that physical properties of a digital chip could be exploited by computer controlled evolution. Subsequently, it has been shown that computer controlled evolution can be used to obtain specific analogue computation in a non-silicon based physical material (liquid crystal). Currently, work is in progress to extend the technique and show that computational processors can be evolved using a range of material systems (e.g. carbon nanotubes, nanoparticles).

The talk reviews past work in the field, discusses the advantages and disadvantages and the challenges that still remain.

Rapporteur

Professor Malcolm Levitt FRS, University of Southampton, UK

Spatial computing: a unifying approach to computational materials

Dr Jacob Beal, BBN Technologies, USA

Abstract

Throughout the course of the 20th century, revolutions in electronic hardware and programming abstractions encouraged us to consider a computer as an abstract box that information flows into and out of. By now, however, we have progressed to the point where the harsh realities of physics and the expanding frontier of non-silicon computing platforms is dissolving that abstraction.  In the emerging new world, a computational system is not a single device, but a space-filling material, with communication, computation, interaction with the physical world distributed throughout.  Continuous space-time abstractions provide a unifying model for approaching such computational materials, even when the devices comprising the material are a mixture of different substrates and different scales in space and time.  I will illustrate this discussion with examples from sensor networks, robotic swarms, and synthetic biology.

Towards the assembly and programming of chemical computers

Professor Lee Cronin, University of Glasgow, UK

Abstract

The construction of complex chemical systems capable of computation is a fantastic challenge due to the ambiguity regarding the definition of the inputs, outputs and processing functions. We could attempt to build systems that are analogous to silicon computers, or use programming techniques that utilise the features of available chemistry set. In my talk I will outline a road map that aims to use a hybrid chemo-robotic system that will program the physical chemical system using a range of different algorithms doing the computation in chemical space rather than in silico.  I will present several new robotic platform that integrates chemistry, algorithms, and sensor systems allowing us to view processes like self-assembly, self-organisation as higher processing functions. From embodied evolution to hybrid chemo-silico computions using molecules as addressable computing elements, we will explore what problems can be solved and how hybrid chemical-silicon computing could define a new computing paradigm.

Heterotic computing

Professor Susan Stepney, University of York, UK

Abstract

Unconventional computers can perform embodied computation that can directly exploit the natural dynamics of the substrate. But such in materio devices are often limited, special purpose machines. To be practically useful, unconventional devices are usually be combined with classical computers or control systems. However, there is currently no established formal way to do this, or to combine different unconventional devices.

I will describe heterotic unconventional computation, a proposed approach that focusses on combinations of unconventional and conventional devices. I will outline a framework suitable for combining diverse unconventional computational devices in a way that respects the intrinsic computational power of each, whilst yielding a hybrid device that is capable of more than the sum of its parts. I will give some potential examples of how this framework could be populated with specific unconventional computing substrates.

Rapporteur

Professor Sir Tony Hoare FREng FRS, Microsoft Research Ltd, Cambridge, UK

Theoretical frameworks

Professor Samson Abramsky FRS, University of Oxford, UK