-
Neuromorphic quadratic programming for efficient and scalable model predictive control
Authors:
Ashish Rao Mangalore,
Gabriel Andres Fonseca Guerra,
Sumedh R. Risbud,
Philipp Stratmann,
Andreas Wild
Abstract:
Applications in robotics or other size-, weight- and power-constrained autonomous systems at the edge often require real-time and low-energy solutions to large optimization problems. Event-based and memory-integrated neuromorphic architectures promise to solve such optimization problems with superior energy efficiency and performance compared to conventional von Neumann architectures. Here, we pre…
▽ More
Applications in robotics or other size-, weight- and power-constrained autonomous systems at the edge often require real-time and low-energy solutions to large optimization problems. Event-based and memory-integrated neuromorphic architectures promise to solve such optimization problems with superior energy efficiency and performance compared to conventional von Neumann architectures. Here, we present a method to solve convex continuous optimization problems with quadratic cost functions and linear constraints on Intel's scalable neuromorphic research chip Loihi 2. When applied to model predictive control (MPC) problems for the quadruped robotic platform ANYmal, this method achieves over two orders of magnitude reduction in combined energy-delay product compared to the state-of-the-art solver, OSQP, on (edge) CPUs and GPUs with solution times under ten milliseconds for various problem sizes. These results demonstrate the benefit of non-von-Neumann architectures for robotic control applications.
△ Less
Submitted 19 June, 2024; v1 submitted 26 January, 2024;
originally announced January 2024.
-
Optical valley separation in two-dimensional semimetals with tilted Dirac cones
Authors:
A. Wild,
E. Mariani,
M. E. Portnoi
Abstract:
Two-dimensional semimetals with tilted Dirac cones in the electronic band structure are shown to exhibit spatial separation of carriers belonging to different valleys under illumination. In stark contrast to gapped Dirac materials this optovalleytronic phenomenon occurs in systems with intact inversion and time-reversal symmetry that host massless Dirac cones in the band structure, thereby retaini…
▽ More
Two-dimensional semimetals with tilted Dirac cones in the electronic band structure are shown to exhibit spatial separation of carriers belonging to different valleys under illumination. In stark contrast to gapped Dirac materials this optovalleytronic phenomenon occurs in systems with intact inversion and time-reversal symmetry that host massless Dirac cones in the band structure, thereby retaining the exceptional graphene-like transport properties. As a result we demonstrate that optical valley separation is possible at arbitrarily low photon frequencies including the deep infrared and terahertz regimes with full gate tunability via Pauli blocking. As a specific example of our theory, we demonstrate tunable valley separation in the proposed two-dimensional tilted Dirac cone semimetal 8-$Pmmn$ borophene for incident infrared photons at room temperature.
△ Less
Submitted 11 January, 2023;
originally announced January 2023.
-
Optical absorption in two-dimensional materials with tilted Dirac cones
Authors:
A. Wild,
E. Mariani,
M. E. Portnoi
Abstract:
The interband optical absorption of linearly polarised light by two-dimensional (2D) semimetals hosting tilted and anisotropic Dirac cones in the bandstructure is analysed theoretically. Super-critically tilted (type-II) Dirac cones are characterised by an absorption that is highly dependent on the incident photon polarisation and frequency, and is tunable by changing the Fermi level with a back-g…
▽ More
The interband optical absorption of linearly polarised light by two-dimensional (2D) semimetals hosting tilted and anisotropic Dirac cones in the bandstructure is analysed theoretically. Super-critically tilted (type-II) Dirac cones are characterised by an absorption that is highly dependent on the incident photon polarisation and frequency, and is tunable by changing the Fermi level with a back-gate voltage. Type-II Dirac cones exhibit open Fermi surfaces and large regions of the Brillouin zone where the valence and conduction bands sit either above or below the Fermi level. As a consequence, unlike their sub-critically tilted (type-I) counterparts, type-II Dirac cones have many states that are Pauli blocked even when the Fermi level is tuned to the level crossing point. We analyse the interplay of the tilt parameter with the Fermi velocity anisotropy, demonstrating that the optical response of a Dirac cone cannot be described by its tilt alone. As a special case of our general theory we discuss the proposed 2D type-I semimetal 8-$Pmmn$ Borophene. Guided by our in-depth analytics we develop an optical recipe to fully characterise the tilt and Fermi velocity anisotropy of any 2D tilted Dirac cone solely from its absorption spectrum. We expect our work to encourage Dirac cone engineering as a major route to create gate-tunable thin-film polarisers.
△ Less
Submitted 8 June, 2022; v1 submitted 1 December, 2021;
originally announced December 2021.
-
Map** borophene onto graphene: Quasi-exact solutions for guiding potentials in tilted Dirac cones
Authors:
R. A. Ng,
A. Wild,
M. E. Portnoi,
R. R. Hartmann
Abstract:
We show that if the solutions to the (2+1)-dimensional massless Dirac equation for a given 1D potential are known, then they can be used to obtain the eigenvalues and eigenfunctions for the same potential, orientated at an arbitrary angle, in a tilted anisotropic 2D Dirac material. This simple set of transformations enables all the exact and quasi-exact solutions associated with 1D quantum wells i…
▽ More
We show that if the solutions to the (2+1)-dimensional massless Dirac equation for a given 1D potential are known, then they can be used to obtain the eigenvalues and eigenfunctions for the same potential, orientated at an arbitrary angle, in a tilted anisotropic 2D Dirac material. This simple set of transformations enables all the exact and quasi-exact solutions associated with 1D quantum wells in graphene to be applied to the confinement problem in tilted Dirac materials such as borophene. We also show that smooth electron waveguides in tilted Dirac materials can be used to manipulate the degree of valley polarization of quasiparticles travelling along a particular direction of the channel. We examine the particular case of the hyperbolic secant potential to model realistic top-gated structures for valleytronic applications.
△ Less
Submitted 21 November, 2021;
originally announced November 2021.
-
Clinica: an open source software platform for reproducible clinical neuroscience studies
Authors:
Alexandre Routier,
Ninon Burgos,
Mauricio Díaz,
Michael Bacci,
Simona Bottani,
Omar El-Rifai,
Sabrina Fontanella,
Pietro Gori,
Jérémy Guillon,
Alexis Guyot,
Ravi Hassanaly,
Thomas Jacquemont,
Pascal Lu,
Arnaud Marcoux,
Tristan Moreau,
Jorge Samper-González,
Marc Teichmann,
Elina Thibeau--Sutre,
Ghislain Vaillant,
Junhao Wen,
Adam Wild,
Marie-Odile Habert,
Stanley Durrleman,
Olivier Colliot
Abstract:
We present Clinica (www.clinica.run), an open-source software platform designed to make clinical neuroscience studies easier and more reproducible. Clinica aims for researchers to i) spend less time on data management and processing, ii) perform reproducible evaluations of their methods, and iii) easily share data and results within their institution and with external collaborators. The core of Cl…
▽ More
We present Clinica (www.clinica.run), an open-source software platform designed to make clinical neuroscience studies easier and more reproducible. Clinica aims for researchers to i) spend less time on data management and processing, ii) perform reproducible evaluations of their methods, and iii) easily share data and results within their institution and with external collaborators. The core of Clinica is a set of automatic pipelines for processing and analysis of multimodal neuroimaging data (currently, T1-weighted MRI, diffusion MRI and PET data), as well as tools for statistics, machine learning and deep learning. It relies on the brain imaging data structure (BIDS) for the organization of raw neuroimaging datasets and on established tools written by the community to build its pipelines. It also provides converters of public neuroimaging datasets to BIDS (currently ADNI, AIBL, OASIS and NIFD). Processed data include image-valued scalar fields (e.g. tissue probability maps), meshes, surface-based scalar fields (e.g. cortical thickness maps) or scalar outputs (e.g. regional averages). These data follow the ClinicA Processed Structure (CAPS) format which shares the same philosophy as BIDS. Consistent organization of raw and processed neuroimaging files facilitates the execution of single pipelines and of sequences of pipelines, as well as the integration of processed data into statistics or machine learning frameworks. The target audience of Clinica is neuroscientists or clinicians conducting clinical neuroscience studies involving multimodal imaging, and researchers develo** advanced machine learning algorithms applied to neuroimaging data.
△ Less
Submitted 21 July, 2021;
originally announced July 2021.
-
A Long Short-Term Memory for AI Applications in Spike-based Neuromorphic Hardware
Authors:
Philipp Plank,
Arjun Rao,
Andreas Wild,
Wolfgang Maass
Abstract:
Spike-based neuromorphic hardware holds the promise to provide more energy efficient implementations of Deep Neural Networks (DNNs) than standard hardware such as GPUs. But this requires to understand how DNNs can be emulated in an event-based sparse firing regime, since otherwise the energy-advantage gets lost. In particular, DNNs that solve sequence processing tasks typically employ Long Short-T…
▽ More
Spike-based neuromorphic hardware holds the promise to provide more energy efficient implementations of Deep Neural Networks (DNNs) than standard hardware such as GPUs. But this requires to understand how DNNs can be emulated in an event-based sparse firing regime, since otherwise the energy-advantage gets lost. In particular, DNNs that solve sequence processing tasks typically employ Long Short-Term Memory (LSTM) units that are hard to emulate with few spikes. We show that a facet of many biological neurons, slow after-hyperpolarizing (AHP) currents after each spike, provides an efficient solution. AHP-currents can easily be implemented in neuromorphic hardware that supports multi-compartment neuron models, such as Intel's Loihi chip. Filter approximation theory explains why AHP-neurons can emulate the function of LSTM units. This yields a highly energy-efficient approach to time series classification. Furthermore it provides the basis for implementing with very sparse firing an important class of large DNNs that extract relations between words and sentences in a text in order to answer questions about the text.
△ Less
Submitted 7 November, 2021; v1 submitted 8 July, 2021;
originally announced July 2021.
-
Automatic quality control of brain T1-weighted magnetic resonance images for a clinical data warehouse
Authors:
Simona Bottani,
Ninon Burgos,
Aurélien Maire,
Adam Wild,
Sebastian Ströer,
Didier Dormont,
Olivier Colliot
Abstract:
Many studies on machine learning (ML) for computer-aided diagnosis have so far been mostly restricted to high-quality research data. Clinical data warehouses, gathering routine examinations from hospitals, offer great promises for training and validation of ML models in a realistic setting. However, the use of such clinical data warehouses requires quality control (QC) tools. Visual QC by experts…
▽ More
Many studies on machine learning (ML) for computer-aided diagnosis have so far been mostly restricted to high-quality research data. Clinical data warehouses, gathering routine examinations from hospitals, offer great promises for training and validation of ML models in a realistic setting. However, the use of such clinical data warehouses requires quality control (QC) tools. Visual QC by experts is time-consuming and does not scale to large datasets. In this paper, we propose a convolutional neural network (CNN) for the automatic QC of 3D T1-weighted brain MRI for a large heterogeneous clinical data warehouse. To that purpose, we used the data warehouse of the hospitals of the Greater Paris area (Assistance Publique-Hôpitaux de Paris [AP-HP]). Specifically, the objectives were: 1) to identify images which are not proper T1-weighted brain MRIs; 2) to identify acquisitions for which gadolinium was injected; 3) to rate the overall image quality. We used 5000 images for training and validation and a separate set of 500 images for testing. In order to train/validate the CNN, the data were annotated by two trained raters according to a visual QC protocol that we specifically designed for application in the setting of a data warehouse. For objectives 1 and 2, our approach achieved excellent accuracy (balanced accuracy and F1-score \textgreater 90\%), similar to the human raters. For objective 3, the performance was good but substantially lower than that of human raters. Nevertheless, the automatic approach accurately identified (balanced accuracy and F1-score \textgreater 80\%) low quality images, which would typically need to be excluded. Overall, our approach shall be useful for exploiting hospital data warehouses in medical image computing.
△ Less
Submitted 16 April, 2021;
originally announced April 2021.
-
NxTF: An API and Compiler for Deep Spiking Neural Networks on Intel Loihi
Authors:
Bodo Rueckauer,
Connor Bybee,
Ralf Goettsche,
Yashwardhan Singh,
Joyesh Mishra,
Andreas Wild
Abstract:
Spiking Neural Networks (SNNs) are a promising paradigm for efficient event-driven processing of spatio-temporally sparse data streams. SNNs have inspired the design and can take advantage of the emerging class of neuromorphic processors like Intel Loihi. These novel hardware architectures expose a variety of constraints that affect firmware, compiler and algorithm development alike. To enable rap…
▽ More
Spiking Neural Networks (SNNs) are a promising paradigm for efficient event-driven processing of spatio-temporally sparse data streams. SNNs have inspired the design and can take advantage of the emerging class of neuromorphic processors like Intel Loihi. These novel hardware architectures expose a variety of constraints that affect firmware, compiler and algorithm development alike. To enable rapid and flexible development of SNN algorithms on Loihi, we developed NxTF: a programming interface derived from Keras and compiler optimized for map** deep convolutional SNNs to the multi-core Intel Loihi architecture. We evaluate NxTF on DNNs trained directly on spikes as well as models converted from traditional DNNs, processing both sparse event-based and dense frame-based data sets. Further, we assess the effectiveness of the compiler to distribute models across a large number of cores and to compress models by exploiting Loihi's weight sharing features. Finally, we evaluate model accuracy, energy and time to solution compared to other architectures. The compiler achieves near optimal resource utilization of 80% across 16 Loihi chips for a 28-layer, 4M parameter MobileNet model with input size 128x128. In addition, we report the lowest error rate of 8.52% for the CIFAR-10 dataset on neuromorphic hardware, using an off-the-shelf MobileNet.
△ Less
Submitted 11 January, 2021;
originally announced January 2021.
-
Neuromorphic Nearest-Neighbor Search Using Intel's Pohoiki Springs
Authors:
E. Paxon Frady,
Garrick Orchard,
David Florey,
Nabil Imam,
Ruokun Liu,
Joyesh Mishra,
Jonathan Tse,
Andreas Wild,
Friedrich T. Sommer,
Mike Davies
Abstract:
Neuromorphic computing applies insights from neuroscience to uncover innovations in computing technology. In the brain, billions of interconnected neurons perform rapid computations at extremely low energy levels by leveraging properties that are foreign to conventional computing systems, such as temporal spiking codes and finely parallelized processing units integrating both memory and computatio…
▽ More
Neuromorphic computing applies insights from neuroscience to uncover innovations in computing technology. In the brain, billions of interconnected neurons perform rapid computations at extremely low energy levels by leveraging properties that are foreign to conventional computing systems, such as temporal spiking codes and finely parallelized processing units integrating both memory and computation. Here, we showcase the Pohoiki Springs neuromorphic system, a mesh of 768 interconnected Loihi chips that collectively implement 100 million spiking neurons in silicon. We demonstrate a scalable approximate k-nearest neighbor (k-NN) algorithm for searching large databases that exploits neuromorphic principles. Compared to state-of-the-art conventional CPU-based implementations, we achieve superior latency, index build time, and energy efficiency when evaluated on several standard datasets containing over 1 million high-dimensional patterns. Further, the system supports adding new data points to the indexed database online in O(1) time unlike all but brute force conventional k-NN implementations.
△ Less
Submitted 27 April, 2020;
originally announced April 2020.
-
Verification for Machine Learning, Autonomy, and Neural Networks Survey
Authors:
Weiming Xiang,
Patrick Musau,
Ayana A. Wild,
Diego Manzanas Lopez,
Nathaniel Hamilton,
Xiaodong Yang,
Joel Rosenfeld,
Taylor T. Johnson
Abstract:
This survey presents an overview of verification techniques for autonomous systems, with a focus on safety-critical autonomous cyber-physical systems (CPS) and subcomponents thereof. Autonomy in CPS is enabling by recent advances in artificial intelligence (AI) and machine learning (ML) through approaches such as deep neural networks (DNNs), embedded in so-called learning enabled components (LECs)…
▽ More
This survey presents an overview of verification techniques for autonomous systems, with a focus on safety-critical autonomous cyber-physical systems (CPS) and subcomponents thereof. Autonomy in CPS is enabling by recent advances in artificial intelligence (AI) and machine learning (ML) through approaches such as deep neural networks (DNNs), embedded in so-called learning enabled components (LECs) that accomplish tasks from classification to control. Recently, the formal methods and formal verification community has developed methods to characterize behaviors in these LECs with eventual goals of formally verifying specifications for LECs, and this article presents a survey of many of these recent approaches.
△ Less
Submitted 3 October, 2018;
originally announced October 2018.
-
Cubic Rashba spin-orbit interaction of two-dimensional hole gas in strained-Ge/SiGe quantum well
Authors:
Rai Moriya,
Kentarou Sawano,
Yusuke Hoshi,
Satoru Masubuchi,
Yasuhiro Shiraki,
Andreas Wild,
Christian Neumann,
Gerhard Abstreiter,
Dominique Bougeard,
Takaaki Koga,
Tomoki Machida
Abstract:
The spin-orbit interaction (SOI) of the two-dimensional hole gas in the inversion symmetric semiconductor Ge is studied in a strained-Ge/SiGe quantum well structure. We observed weak anti-localization (WAL) in the magnetoconductivity measurement, revealing that the WAL feature can be fully described by the k-cubic Rashba SOI theory. Furthermore, we demonstrated electric field control of the Rashba…
▽ More
The spin-orbit interaction (SOI) of the two-dimensional hole gas in the inversion symmetric semiconductor Ge is studied in a strained-Ge/SiGe quantum well structure. We observed weak anti-localization (WAL) in the magnetoconductivity measurement, revealing that the WAL feature can be fully described by the k-cubic Rashba SOI theory. Furthermore, we demonstrated electric field control of the Rashba SOI. Our findings reveal that the heavy hole (HH) in strained-Ge is a purely cubic-Rashba-system, which is consistent with the spin angular momentum mj = +-3/2 nature of the HH wave function.
△ Less
Submitted 7 August, 2014; v1 submitted 5 August, 2014;
originally announced August 2014.
-
Sneaking Up On The Criegee Intermediate From Below: Predicted Photoelectron Spectrum Of The CH_2OO^- Anion And W3-F12 Electron Affinity Of CH_2OO
Authors:
Amir Karton,
Marcus Kettner,
Duncan Andrew Wild
Abstract:
High level ab initio calculations were undertaken on the CH$_2$OO anion and neutral species to predict the electron affinity and anion photoelectron spectrum. The electron affinity of CH$_2$OO, 0.567 eV, and barrier height for dissociation of CH$_2$OO$^-$ to O$^-$ and CH$_2$O, 16.5 kJ mol$^{-1}$, are obtained by means of the W3-F12 thermochemical protocol. Two major geometric differences between t…
▽ More
High level ab initio calculations were undertaken on the CH$_2$OO anion and neutral species to predict the electron affinity and anion photoelectron spectrum. The electron affinity of CH$_2$OO, 0.567 eV, and barrier height for dissociation of CH$_2$OO$^-$ to O$^-$ and CH$_2$O, 16.5 kJ mol$^{-1}$, are obtained by means of the W3-F12 thermochemical protocol. Two major geometric differences between the anion and neutral, being the dihedral angle of the terminal hydrogen atoms with respect to C-O-O plane, and the O-O bond length, are reflected in the predicted spectrum as pronounced vibrational progressions.
△ Less
Submitted 11 September, 2013; v1 submitted 10 September, 2013;
originally announced September 2013.
-
Few electron double quantum dot in an isotopically purified $^{28}$Si quantum well
Authors:
Andreas Wild,
Johannes Kierig,
Jürgen Sailer,
Joel Ager III,
Eugene Haller,
Gerhard Abstreiter,
Stefan Ludwig,
Dominique Bougeard
Abstract:
We present a few electron double quantum dot (QD) device defined in an isotopically purified $^{28}$Si quantum well (QW). An electron mobility of $5.5 \cdot 10^4 cm^2(Vs)^{-1}$ is observed in the QW which is the highest mobility ever reported for a 2D electron system in $^{28}$Si. The residual concentration of $^{29}$Si nuclei in the $^{28}$Si QW is lower than $10^{3} ppm$, at the verge where the…
▽ More
We present a few electron double quantum dot (QD) device defined in an isotopically purified $^{28}$Si quantum well (QW). An electron mobility of $5.5 \cdot 10^4 cm^2(Vs)^{-1}$ is observed in the QW which is the highest mobility ever reported for a 2D electron system in $^{28}$Si. The residual concentration of $^{29}$Si nuclei in the $^{28}$Si QW is lower than $10^{3} ppm$, at the verge where the hyperfine interaction is theoretically no longer expected to dominantly limit the $T_{2}$ spin dephasing time. We also demonstrate a complete suppression of hysteretic gate behavior and charge noise using a negatively biased global top gate.
△ Less
Submitted 6 April, 2012; v1 submitted 15 February, 2012;
originally announced February 2012.
-
Quantum Hall Resistance Overshoot in 2-Dimensional Electron Gases - Theory and Experiment
Authors:
J. Sailer,
A. Wild,
V. Lang,
A. Siddiki,
D. Bougeard
Abstract:
We present a systematical experimental investigation of an unusual transport phenomenon observed in two dimensional electron gases in Si/SiGe heterostructures under integer quantum Hall effect (IQHE) conditions. This phenomenon emerges under specific experimental conditions and in different material systems. It is commonly referred to as Hall resistance overshoot, however, lacks a consistent expla…
▽ More
We present a systematical experimental investigation of an unusual transport phenomenon observed in two dimensional electron gases in Si/SiGe heterostructures under integer quantum Hall effect (IQHE) conditions. This phenomenon emerges under specific experimental conditions and in different material systems. It is commonly referred to as Hall resistance overshoot, however, lacks a consistent explanation so far. Based on our experimental findings we are able to develop a model that accounts for all of our observations in the framework of a screening theory for the IQHE. Within this model the origin of the overshoot is attributed to a transport regime where current is confined to co-existing evanescent incompressible strips of different filling factors.
△ Less
Submitted 15 July, 2010;
originally announced July 2010.
-
Electrostatically defined Quantum Dots in a Si/SiGe Heterostructure
Authors:
A. Wild,
J. Sailer,
J. Nützel,
G. Abstreiter,
S. Ludwig,
D. Bougeard
Abstract:
We present an electrostatically defined few-electron double quantum dot (QD) realized in a molecular beam epitaxy grown Si/SiGe heterostructure. Transport and charge spectroscopy with an additional QD as well as pulsed-gate measurements are demonstrated. We discuss technological challenges specific for silicon-based heterostructures and the effect of a comparably large effective electron mass on t…
▽ More
We present an electrostatically defined few-electron double quantum dot (QD) realized in a molecular beam epitaxy grown Si/SiGe heterostructure. Transport and charge spectroscopy with an additional QD as well as pulsed-gate measurements are demonstrated. We discuss technological challenges specific for silicon-based heterostructures and the effect of a comparably large effective electron mass on transport properties and tunability of the double QD. Charge noise, which might be intrinsically induced due to strain-engineering is proven not to affect the stable operation of our device as a spin qubit. Our results promise the suitability of electrostatically defined QDs in Si/SiGe heterostructures for quantum information processing.
△ Less
Submitted 14 July, 2010;
originally announced July 2010.
