News and posts

MemComputing: leveraging physics to compute efficiently

Time: Tuesday, October 9th, 15:00
Speaker: Massimiliano DI VENTRA, UCSD
It is well known that physical phenomena may be of great help in computing some difficult problems efficiently. A typical example is prime factorization that may be solved in polynomial time by exploiting quantum entanglement on a quantum computer. There are, however, other types of (non-quantum) physical properties that one may leverage to compute efficiently a wide range of hard problems. In this talk I will discuss how to employ one such property, memory (time non-locality), in a novel physics-based approach to computation: Memcomputing [1, 2, 3, 4]. As examples, I will show the polynomial-time solution of prime factorization, the search version of the subset-sum problem [5], and approximations to the Max-SAT beyond the inapproximability gap [6] using polynomial resources and self-organizing logic gates, namely gates that self-organize to satisfy their logical proposition [5]. I will also show that these machines are described by a Witten-type topological field theory, and they compute via an instantonic phase, implying that they are robust against noise and disorder [7]. The digital memcomputing machines we propose can be efficiently simulated, are scalable and can be easily realized with available nanotechnology components. Work supported in part by MemComputing, Inc. (http://memcpu.com/).
[1] F. L. Traversa and M. Di Ventra, Universal Memcomputing Machines, IEEE Transactions on Neural Networks and Learning Systems 26, 2702 (2015).
[2] M. Di Ventra and Y.V. Pershin, Computing: the Parallel Approach, Nature Physics 9, 200 (2013).
[3] M. Di Ventra and Y.V. Pershin, Just add memory, Scientific American 312, 56 (2015).
[4] M. Di Ventra and F.L. Traversa, Memcomputing: leveraging memory and physics to compute efficiently, J. Appl. Phys. 123, 180901 (2018).
[5] F. L. Traversa and M. Di Ventra, Polynomial-time solution of prime factorization and NP-complete problems with digital memcomputing machines, Chaos: An Interdisciplinary Journal of Nonlinear Science 27, 023107 (2017).
[6] F. L. Traversa, P. Cicotti, F. Sheldon, and M. Di Ventra, Evidence of an exponential speed-up in the solution of hard optimization problems, Complexity 2018, 7982851 (2018).
[7] M. Di Ventra, F. L. Traversa and I.V. Ovchinnikov, Topological field theory and computing with instantons, Annalen der Physik 529,1700123 (2017).

Skyrmions based Neurmorphic Computing

Time: Tuesday, October 9th, 14:00
Speaker: Weisheng ZHAO, Beihang University

Recently, magnetic skyrmion, a swirling topological spin configuration, has been studied as a promising information carrier candidate in future ultra-dense, low-power memory and logic devices for its outstanding merits of nanoscale size, low depinning current density, high motion velocity and particle-like stability etc. One of the most potential applications is to design racetrack memory. One the other hand, we exploit the dynamics of skyrmions to design neuromorphic computing.

Go back

Computing with Spin-Wave Solitons

Time: Tuesday, October 9th, 11:00
Speaker: Ferran MACIA, Universita de Barcelona
Collective magnetic excitations, such as spin waves are attracting a growing interest for their potential applications as memory and communication devices working at high frequency and low power. In particular the use of nanometer scale oscillators could lead to new types information processing, including neuromorphic computing. In this talk we will review some applications of spin-wave patterns created from Spin Torque Oscillators (STO) and their interactions with background oscillations [1-4]. Next, we will focus on solitonic modes that can be created in STO. First I will show that arrays of vortices can be described through the Kuramoto model, having patterns of synchronization resembling the patterns occurring in the brain [5]. Second, I will discuss some advances in the study of magnetic Droplet Solitons and Dynamical Skyrmions, which are magnetic excitations consisting of reversed oscillating spins that are strongly localized and can be controlled through the applied field and the spin current [6-9]. Droplet solitons are multistate oscillators with a hysteretic behavior making them good candidates as building blocks in neuromorphic computing

[1] F. Macià et al. Nanotechnology 22, 095301 (2011)
[2] F. Macià et al. Nanotechnology 25, 045303 (2014)
[3] F.C. Hoppensteadt, US Patent 9,582,695 (2018)
[4] F Hoppensteadt, Biosystems 136, 99-104 (2017)
[5] V. Flovick et al Scientific Reports, 6, 32528 (2016)
[6] S. M. Mohseni et al., Science 339, 1295 (2013)
[7] F. Macià et al. Nature Nanotech. (2014)
[8] J.Hang et al. Scientific Reports 8, 6847, (2018)
[9] N. Statuto et al. Nanotechnology, 29, 325302 (2018)

Mutually synchronized spin Hall nano-oscillator arrays

Time: Tuesday, October 9th, 10:10
Speaker: Johan AKERMAN, Gothenburg University

Spin Hall nano-oscillators (SHNOs) [1] are an emerging class of nano-scopic microwave signal generators with potential for new disruptive applications ranging from microwave signal generation/detection to neuromorphic computing [2,3]. SHNOs are based on an intrinsic magnetodynamic resonance with frequencies in the GHz range, which depends on material parameters, device layout, and external parameters such as magnetic field and drive current. For sufficiently high current densities, the resonance can be driven into a state of coherent auto-oscillation. Through the magnetoresistance of the device, the auto-oscillation can be used to generate a current- and field-tunable microwave voltage.
The auto-oscillation state is highly non-linear in nature, and neighbouring SHNOs can therefore interact with each other and even mutually synchronize, which further increases the power and coherence of the microwave signal [4]. This is important, as the nano-scale volume of the auto-oscillating spin wave mode is susceptible to thermal noise, leading to detrimental phase noise in the microwave signal. I will present resent results on long chains of SHNOs and the first two-dimensional SHNO arrays. We demonstrate robust mutual synchronization in chains or 21 SHNOs with record quality factors of Q=f/df of 30,000. We also demonstrate robust mutual synchronization in two-dimensional arrays of as many as 8 x 8 = 64 SHNOs. We find that the linewidth of these arrays decreases linearly with the number of SHNOs, which enables us to reach Q factors as high as 170,000, i.e. an order of magnitude higher than literature values. Based on these results I will argue that a viable path towards commercial microwave signal generators based on spintronic devices must be based on mutually synchronized SHNO arrays.
The mutual synchronization phenomenon can also be used for ultra-fast pattern matching with potential for speeding up image recognition by orders of magnitude. With the recent rapidly increasing interest in artificial intelligence and neuromorphic computing, mutually synchronized SHNO chains and arrays hence represent a highly attractive emerging technology platform for low-power, and ultrafast non-conventional computing.
References
[1] T. Chen, et al, Proc. IEEE 104, 1919 (2016).
[2] J. Grollier, D. Querlioz, and M. Stiles, Proc. IEEE 104, 2024 (2016).
[3] J. Torrejon et al., Nature 547, 428 (2017)
[4] A. A. Awad et al. Nature Physics 13, 292 (2017).
[5] M. Zahedinejad et al. unpublished (2018).

Go back

Tutorial: Manipulating Magnetic Skyrmions

Time: Tuesday, October 9th, 09:00
Speaker: Axel HOFFMANN, Argonne National Laboratory
Magnetic skyrmions are topologically distinct spin textures that are stabilized by the interplay between applied magnetic fields, magnetic anisotropies, as well as symmetric and antisymmetric exchange interactions. Due to their topology magnetic skyrmions can be stable with quasi-particle like behavior, where they can be manipulated with very low electric currents. This makes them interesting for low-power information technologies, where it is envisioned that data will be encoded in topological charges, instead of electronic charges as in conventional semiconducting devices. In particular, recently there has been a lot of progress stabilizing magnetic skyrmions at room temperature in magnetic heterostructures [1]. This talk will review specific aspects that relate to the manipulation of individual magnetic skyrmions, such as their electrical generation [2,3], motion [4], and dynamical excitations.
This work was supported by the U.S. Department of Energy, Office of Science, Materials Sciences and Engineering Division.

[1] W. Jiang, et al., Phys. Rep. 704, 1 (2017).
[2] W. Jiang, et al., Science 349, 283 (2015).
[3] O. Heinonen, et al., Phys. Rev. B 93, 094407 (2016).
[4] W. Jiang, et al., Nature Phys. 13, 162 (2017).

Tutorial: Computing with magnetic dots and spintronic dynamical systems

Time: Monday, October 8th, 15:00
Speaker: Wolfgang POROD, University of Notre Dame
In this talk, we will discuss alternative approaches to computing. Instead of basing a computer on the manipulation of bits, represented by electronic switches, we will explore the possibility of harnessing spintronic dynamical systems for computation. Specifically, we will consider physical processes and dynamical systems based on coupled magnetic dots, on coupled spin-torque oscillators, and on spin waves. In such dynamical systems, the computational process more closely exploits the underlying physics, which offers the promise of lower power dissipation.

Tutorial: Reinforcement Learning

Time: Monday, October 8th, 10:20
Speaker: Eleni VASILAKI, University of Sheffield
In this tutorial, I will precent the key concepts behind Reinforcement Learning (RL), and an overview of various reinforcement learning frameworks. I will also discuss its link to Deep Learning, and comment on what RL tells us about happiness.

Tutorial: The Landscape of Deep Learning: a Quick Overview

Time: Monday, October 8th, 09:10
Speaker: Teodora PETRISOR, Thales Group
This talk is a walk-through to the basic principles, theoretical concepts and some of the successful models in Artificial Neural Networks today with a particular focus on the so-called Deep Learning paradigm, all from an algorithmic viewpoint. We will also highlight some of the challenges to be addressed for the large-scale industrialization of these methods.

Go back

All-optical switching and brain-inspired concepts for low energy information processing

Time: Monday, October 8th, 14:00
Speaker: Theo RASING, Radboud University

The explosive growth of big data and artificial intelligence offers a huge potential for new digital products and unexplored business models. While data has become an indispensable part of modern society, the sheer amount of data being generated every second is breathtaking, both in its scale and in its growth, while the number of devices generating these data is rapidly expanding. This not only pushes current technologies to their limits, but also that of our energy production: our ICT and data centres already consume around 7% of the world’s electricity production and with the growth rate of ICT-technologies, this energy consumption is rapidly becoming unsustainable. In stark contrast, the human brain, with its intricate architecture combining both processing and storing of information, only consumes about 10 Watt of energy while having a similar capacity as a supercomputer consuming around 10 Megawatt.
We try to develop materials and concepts that mimic the efficiency of the brain by combining local processing and storage, using adaptable physical interactions that can implement learning algorithms. We demonstrate, by modelling, that a reconfigurable and self-learning structure can be achieved, which implements the prototype perceptron model of a neural network based on magneto-optical interactions. Importantly, we show that optimization of synaptic weights is achieved by a global feedback mechanism, such that learning does not rely on external storage or additional optimization schemes. For the experimental realization of adaptive synaptic structures, we choose to use optically controllable magnetization in a thin Co/Pt film1, using circularly polarized picosecond2 pulse trains. The combined stochastic/deterministic nature of all-optical switching in this material2 offers the possibility to continuously vary the magneto-optical Faraday rotation with the number of pulses, yielding the necessary ingredient to realize a perceptron-like structure. First results of such a learning structure will be demonstrated.

1. C.-H. Lambert et al, Science 345, 1337 (2014)
2. R. Medapalli et al, Phys. Rev.B 96, 224421 (2017)

Go back