-
Powering AI at the Edge: A Robust, Memristor-based Binarized Neural Network with Near-Memory Computing and Miniaturized Solar Cell
Authors:
Fadi Jebali,
Atreya Majumdar,
Clément Turck,
Kamel-Eddine Harabi,
Mathieu-Coumba Faye,
Eloi Muhr,
Jean-Pierre Walder,
Oleksandr Bilousov,
Amadeo Michaud,
Elisa Vianello,
Tifenn Hirtzlin,
François Andrieu,
Marc Bocquet,
Stéphane Collin,
Damien Querlioz,
Jean-Michel Portal
Abstract:
Memristor-based neural networks provide an exceptional energy-efficient platform for artificial intelligence (AI), presenting the possibility of self-powered operation when paired with energy harvesters. However, most memristor-based networks rely on analog in-memory computing, necessitating a stable and precise power supply, which is incompatible with the inherently unstable and unreliable energy…
▽ More
Memristor-based neural networks provide an exceptional energy-efficient platform for artificial intelligence (AI), presenting the possibility of self-powered operation when paired with energy harvesters. However, most memristor-based networks rely on analog in-memory computing, necessitating a stable and precise power supply, which is incompatible with the inherently unstable and unreliable energy harvesters. In this work, we fabricated a robust binarized neural network comprising 32,768 memristors, powered by a miniature wide-bandgap solar cell optimized for edge applications. Our circuit employs a resilient digital near-memory computing approach, featuring complementarily programmed memristors and logic-in-sense-amplifier. This design eliminates the need for compensation or calibration, operating effectively under diverse conditions. Under high illumination, the circuit achieves inference performance comparable to that of a lab bench power supply. In low illumination scenarios, it remains functional with slightly reduced accuracy, seamlessly transitioning to an approximate computing mode. Through image classification neural network simulations, we demonstrate that misclassified images under low illumination are primarily difficult-to-classify cases. Our approach lays the groundwork for self-powered AI and the creation of intelligent sensors for various applications in health, safety, and environment monitoring.
△ Less
Submitted 22 May, 2023;
originally announced May 2023.
-
Experimental demonstration of Single-Level and Multi-Level-Cell RRAM-based In-Memory Computing with up to 16 parallel operations
Authors:
E. Esmanhotto,
T. Hirtzlin,
N. Castellani,
S. Martin,
B. Giraud,
F. Andrieu,
J. F. Nodin,
D. Querlioz,
J-M. Portal,
E. Vianello
Abstract:
Crossbar arrays of resistive memories (RRAM) hold the promise of enabling In-Memory Computing (IMC), but essential challenges due to the impact of device imperfection and device endurance have yet to be overcome. In this work, we demonstrate experimentally an RRAM-based IMC logic concept with strong resilience to RRAM variability, even after one million endurance cycles. Our work relies on a gener…
▽ More
Crossbar arrays of resistive memories (RRAM) hold the promise of enabling In-Memory Computing (IMC), but essential challenges due to the impact of device imperfection and device endurance have yet to be overcome. In this work, we demonstrate experimentally an RRAM-based IMC logic concept with strong resilience to RRAM variability, even after one million endurance cycles. Our work relies on a generalization of the concept of in-memory Scouting Logic, and we demonstrate it experimentally with up to 16 parallel devices (operands), a new milestone for RRAM in-memory logic. Moreover, we combine IMC with Multi-Level-Cell programming and demonstrate experimentally, for the first time, an IMC RRAM-based MLC 2-bit adder.
△ Less
Submitted 3 March, 2022;
originally announced March 2022.
-
Hardware calibrated learning to compensate heterogeneity in analog RRAM-based Spiking Neural Networks
Authors:
Filippo Moro,
E. Esmanhotto,
T. Hirtzlin,
N. Castellani,
A. Trabelsi,
T. Dalgaty,
G. Molas,
F. Andrieu,
S. Brivio,
S. Spiga,
G. Indiveri,
M. Payvand,
E. Vianello
Abstract:
Spiking Neural Networks (SNNs) can unleash the full power of analog Resistive Random Access Memories (RRAMs) based circuits for low power signal processing. Their inherent computational sparsity naturally results in energy efficiency benefits. The main challenge implementing robust SNNs is the intrinsic variability (heterogeneity) of both analog CMOS circuits and RRAM technology. In this work, we…
▽ More
Spiking Neural Networks (SNNs) can unleash the full power of analog Resistive Random Access Memories (RRAMs) based circuits for low power signal processing. Their inherent computational sparsity naturally results in energy efficiency benefits. The main challenge implementing robust SNNs is the intrinsic variability (heterogeneity) of both analog CMOS circuits and RRAM technology. In this work, we assessed the performance and variability of RRAM-based neuromorphic circuits that were designed and fabricated using a 130\,nm technology node. Based on these results, we propose a Neuromorphic Hardware Calibrated (NHC) SNN, where the learning circuits are calibrated on the measured data. We show that by taking into account the measured heterogeneity characteristics in the off-chip learning phase, the NHC SNN self-corrects its hardware non-idealities and learns to solve benchmark tasks with high accuracy. This work demonstrates how to cope with the heterogeneity of neurons and synapses for increasing classification accuracy in temporal tasks.
△ Less
Submitted 10 February, 2022;
originally announced February 2022.
-
Ultra-High-density 3D vertical RRAM with stacked JunctionLess nanowires for In-Memory-Computing applications
Authors:
M. Ezzadeen,
D. Bosch,
B. Giraud,
S. Barraud,
J. -P. Noel,
D. Lattard,
J. Lacord,
J. -M. Portal,
F. Andrieu
Abstract:
The Von-Neumann bottleneck is a clear limitation for data-intensive applications, bringing in-memory computing (IMC) solutions to the fore. Since large data sets are usually stored in nonvolatile memory (NVM), various solutions have been proposed based on emerging memories, such as OxRAM, that rely mainly on area hungry, one transistor (1T) one OxRAM (1R) bit-cell. To tackle this area issue, while…
▽ More
The Von-Neumann bottleneck is a clear limitation for data-intensive applications, bringing in-memory computing (IMC) solutions to the fore. Since large data sets are usually stored in nonvolatile memory (NVM), various solutions have been proposed based on emerging memories, such as OxRAM, that rely mainly on area hungry, one transistor (1T) one OxRAM (1R) bit-cell. To tackle this area issue, while kee** the programming control provided by 1T1R bit-cell, we propose to combine gate-all-around stacked junctionless nanowires (1JL) and OxRAM (1R) technology to create a 3-D memory pillar with ultrahigh density. Nanowire junctionless transistors have been fabricated, characterized, and simulated to define current conditions for the whole pillar. Finally, based on Simulation Program with Integrated Circuit Emphasis (SPICE) simulations, we demonstrated successfully scouting logic operations up to three-pillar layers, with one operand per layer.
△ Less
Submitted 30 November, 2020;
originally announced December 2020.
-
Laser Processing For 3D Junctionless Transistor Fabrication
Authors:
D. Bosch,
P. Acosta Alba,
S. Kerdiles,
V. Benevent,
C. Perrot,
J. Lassarre,
J. Richy,
J. Lacord,
B. Sklenard,
L. Brunet,
P. Batude,
C. Fenouillet-Beranger,
D. Lattard,
J. P. Colinge,
F. Balestra,
F. Andrieu
Abstract:
To take fully advantage of Junctionless transistor (JLT) low-cost and low-temperature features we investigate a 475 degC process to create onto a wafer a thin poly-Si layer on insulator. We fabricated a 13nm doped (Phosphorous, 1E19 at/cm3) poly-silicon film featuring excellent roughness values (Rmax= 1.6nm and RMS=0.2nm). Guidelines for grain size optimization using nanosecond (ns) laser annealin…
▽ More
To take fully advantage of Junctionless transistor (JLT) low-cost and low-temperature features we investigate a 475 degC process to create onto a wafer a thin poly-Si layer on insulator. We fabricated a 13nm doped (Phosphorous, 1E19 at/cm3) poly-silicon film featuring excellent roughness values (Rmax= 1.6nm and RMS=0.2nm). Guidelines for grain size optimization using nanosecond (ns) laser annealing are given.
△ Less
Submitted 30 November, 2020;
originally announced November 2020.
-
Constraints on effusive cryovolcanic eruptions on Europa using topography obtained from Galileo images
Authors:
Elodie Lesage,
Frédéric Schmidt,
François Andrieu,
Hélène Massol
Abstract:
Images of Europa's surface taken by the Galileo Solid State Imager (SSI) show smooth features measuring a few kilometers, potentially resulting from eruptions of low-viscosity material such as liquid cryomagma. We estimated the volume of four of these smooth features by producing digital elevation models (DEMs) of four Galileo/SSI images. We used the shape-from-shading technique with special care…
▽ More
Images of Europa's surface taken by the Galileo Solid State Imager (SSI) show smooth features measuring a few kilometers, potentially resulting from eruptions of low-viscosity material such as liquid cryomagma. We estimated the volume of four of these smooth features by producing digital elevation models (DEMs) of four Galileo/SSI images. We used the shape-from-shading technique with special care to estimate the uncertainties on the produced DEMs and estimated feature volumes to be between ($5.7*10^{7}$ m$^{3}$ and ($2.7*10^{8}$ m$^{3}$. We discussed the implications for putative sub-surface liquid reservoir dimensions in the case of eruptions induced from freezing reservoirs. Our previous cryovolcanic eruption model was improved by considering a cycle of cryomagma freezing and effusion and by estimating the vaporized cryolava fraction once cryolava spreads onto Europa's surface. Our results show that the cryomagma reservoirs would have to be relatively large to generate these smooth features (1 to 100 km$^{3}$ if the flow features result from a single eruption, and 0.4 to 60 km$^{3}$ for the full lifetime of a reservoir generating cyclic eruptions). The two future missions JUICE (ESA) and Europa Clipper (NASA) should reach Europa during the late 2020s. They shall give more information on those putative cryovolcanic regions which appear as interesting targets that could provide a better understanding of the material exchanges between the surface, sub-surface and ocean of Europa.
△ Less
Submitted 24 February, 2021; v1 submitted 29 September, 2020;
originally announced September 2020.
-
VIRTIS-H observations of comet 67P's dust coma: spectral properties and color temperature variability with phase and elevation
Authors:
D. Bockelée-Morvan,
C. Leyrat,
S. Erard,
F. Andrieu,
F. Capaccioni,
G. Filacchione,
P. H. Hasselmann,
J. Crovisier,
P. Drossart,
G. Arnold,
M. Ciarniello,
D. Kappel,
A. Longobardo,
M. -T. Capria,
M. C. De Sanctis,
G. Rinaldi,
F. Taylor
Abstract:
We analyze 2-5 micrometre spectroscopic observations of the dust coma of comet 67P/Churyumov-Gerasimenko obtained with the VIRTIS-H instrument onboard Rosetta from 3 June to 29 October 2015 at heliocentric distances r_h = 1.24-1.55 AU. The 2-2.5 micrometre color, bolometric albedo, and color temperature are measured using spectral fitting. Data obtained at alpha = 90° solar phase angle show an inc…
▽ More
We analyze 2-5 micrometre spectroscopic observations of the dust coma of comet 67P/Churyumov-Gerasimenko obtained with the VIRTIS-H instrument onboard Rosetta from 3 June to 29 October 2015 at heliocentric distances r_h = 1.24-1.55 AU. The 2-2.5 micrometre color, bolometric albedo, and color temperature are measured using spectral fitting. Data obtained at alpha = 90° solar phase angle show an increase of the bolometric albedo (0.05 to 0.14) with increasing altitude (0.5 to 8 km), accompanied by a possible marginal decrease of the color and color temperature. Possible explanations include the presence in the inner coma of dark particles on ballistic trajectories, and radial changes in particle composition. In the phase angle range 50-120°, phase reddening is significant (0.031 %/100 nm/°), for a mean color of 2 %/100 nm at alpha = 90°, that can be related to the roughness of the dust particles. Moreover, a decrease of the color temperature with decreasing phase angle is also observed at a rate of ~ 0.3 K/°, consistent with the presence of large porous particles, with low thermal inertia, and showing a significant day-to-night temperature contrast. Comparing data acquired at fixed phase angle (alpha = 90°), a 20% increase of the bolometric albedo is observed near perihelion. Heliocentric variations of the dust color are not significant in the analyzed time period. Measured color temperatures are varying from 260 to 320 K, and follow a r^0.6 variation in the r_h = 1.24-1.5 AU range, close to the expected r_h^0.5 value.
△ Less
Submitted 10 January, 2019;
originally announced January 2019.
-
Ice state evolution during spring in Richardson crater, Mars
Authors:
François Andrieu,
Frédéric Schmidt,
Sylvain Douté,
Eric Chassefière
Abstract:
The Martian climate is governed by an annual cycle, that results in the condensation of CO$_{2}$ ice during winter, up to a meter thick at the pole and thousands of kilometers in extension. Water and dust may be trapped during the condensation and freed during the sublimation. In addition, ice may be translucent or granular depending on the deposition process (snow vs direct condensation), anneali…
▽ More
The Martian climate is governed by an annual cycle, that results in the condensation of CO$_{2}$ ice during winter, up to a meter thick at the pole and thousands of kilometers in extension. Water and dust may be trapped during the condensation and freed during the sublimation. In addition, ice may be translucent or granular depending on the deposition process (snow vs direct condensation), annealing efficiency, and dust sinking process. The determination of ice translucency is of particular interest to confirm or reject the cold jet model (also known as Kieffer model).
This work is focused on the dune field of Richardson Crater in which strong interactions between the water, dust and CO$_{2}$ cycles are observed. We analyzed CRISM hyperspectral images in the near IR using radiative transfer model inversion. We demonstrate that among the states of CO$_{2}$ ice, the translucent state is observed most frequently. The monitoring of surface characteristics shows a decrease in the thickness of the ice during the spring consistently with climate models simulations. We estimate a very low dust content of a few ppmv into the CO$_{2}$ ice, consistent with the formation scenario of cold jets. The water impurities is around 0.1\%v, almost stable during the spring, suggesting a water escape from the surface of subliming CO$_{2}$ ice layer. The water ice grain size varies in a range 1 to 50 microns. From these results, we propose the following new mechanism of small water ice grain suspension: as a cold jet occurs, water ice grains of various sizes are lifted from the surface. These jets happen during daytime, when the general upward gas flux from the subliming CO$_{2}$ ice layer is strong enough to carry the smaller grains, while the bigger fall back on the CO$_{2}$ ice layer. The smaller water grains are carried away and integrated to the general atmospheric circulation.
△ Less
Submitted 19 June, 2018;
originally announced June 2018.
-
Formation of recurring slope lineae on Mars by rarefied gas-triggered granular flows
Authors:
F. Schmidt,
F. Andrieu,
F. Costard,
M. Kocifaj,
A. Meresescu
Abstract:
Active dark flows known as recurring slope lineae have been observed on the warmest slopes of equatorial Mars. The morphology, composition and seasonality of the lineae suggest a role of liquid water in their formation. However, internal and atmospheric sources of water appear to be insufficient to sustain the observed slope activity. Experimental evidence suggests that under the low atmospheric p…
▽ More
Active dark flows known as recurring slope lineae have been observed on the warmest slopes of equatorial Mars. The morphology, composition and seasonality of the lineae suggest a role of liquid water in their formation. However, internal and atmospheric sources of water appear to be insufficient to sustain the observed slope activity. Experimental evidence suggests that under the low atmospheric pressure at the surface of Mars, gas can flow upwards through porous Martian soil due to thermal creep under surface regions heated by the Sun, and disturb small particles. Here we present numerical simulations to demonstrate that such a dry process involving the pum** of rarefied gas in the Martian soil due to temperature contrasts can explain the formation of the recurring slope lineae. In our simulations, solar irradiation followed by shadow significantly reduces the angle of repose due to the resulting temporary temperature gradients over shaded terrain, and leads to flow at intermediate slope angles. The simulated flow locations are consistent with observed recurring slope lineae that initiate in rough and bouldered terrains with local shadows over the soil. We suggest that this dry avalanche process can explain the formation of the recurring slope lineae on Mars without requiring liquid water or CO2 frost activity.
△ Less
Submitted 14 February, 2018;
originally announced February 2018.
-
VESPA: a community-driven Virtual Observatory in Planetary Science
Authors:
S. Erard,
B. Cecconi,
P. Le Sidaner,
A. P. Rossi,
T. Capria,
B. Schmitt,
V. Génot,
N. André,
A. C. Vandaele,
M. Scherf,
R. Hueso,
A. Määttänen,
W. Thuillot,
B. Carry,
N. Achilleos,
C. Marmo,
O. Santolik,
K. Benson,
P. Fernique,
L. Beigbeder,
E. Millour,
B. Rousseau,
F. Andrieu,
C. Chauvin,
M. Minin
, et al. (17 additional authors not shown)
Abstract:
The VESPA data access system focuses on applying Virtual Observatory (VO) standards and tools to Planetary Science. Building on a previous EC-funded Europlanet program, it has reached maturity during the first year of a new Europlanet 2020 program (started in 2015 for 4 years). The infrastructure has been upgraded to handle many fields of Solar System studies, with a focus both on users and data p…
▽ More
The VESPA data access system focuses on applying Virtual Observatory (VO) standards and tools to Planetary Science. Building on a previous EC-funded Europlanet program, it has reached maturity during the first year of a new Europlanet 2020 program (started in 2015 for 4 years). The infrastructure has been upgraded to handle many fields of Solar System studies, with a focus both on users and data providers. This paper describes the broad lines of the current VESPA infrastructure as seen by a potential user, and provides examples of real use cases in several thematic areas. These use cases are also intended to identify hints for future developments and adaptations of VO tools to Planetary Science.
△ Less
Submitted 14 June, 2017; v1 submitted 26 May, 2017;
originally announced May 2017.
-
Radiative transfer model for contaminated slabs : experimental validations
Authors:
François Andrieu,
Frédéric Schmidt,
Bernard Schmitt,
Sylvain Douté,
Olivier Brissaud
Abstract:
This article presents a set of spectro-goniometric measurements of different water ice samples and the comparison with an approximated radiative transfer model. The experiments were done using the spectro-radiogoniometer described in Brissaud et al. (2004). The radiative transfer model assumes an isotropization of the flux after the second interface and is fully described in Andrieu et al. (2015).…
▽ More
This article presents a set of spectro-goniometric measurements of different water ice samples and the comparison with an approximated radiative transfer model. The experiments were done using the spectro-radiogoniometer described in Brissaud et al. (2004). The radiative transfer model assumes an isotropization of the flux after the second interface and is fully described in Andrieu et al. (2015). Two kind of experiments were conducted. First, the specular spot was closely investigated, at high angular resolution, at the wavelength of $1.5\,\mbox{μm}$, where ice behaves as a very absorbing media. Second, the bidirectional reflectance was sampled at various geometries, including low phase angles on 61 wavelengths ranging from $0.8\,\mbox{μm}$ to $2.0\,\mbox{μm}$.
In order to validate the model, we made a qualitative test to demonstrate the relative isotropization of the flux. We also conducted quantitative assessments by using a bayesian inversion method in order to estimate the parameters (e.g. sample thickness, surface roughness) from the radiative measurements only. A simple comparison between the retrieved parameters and the direct independent measurements allowed us to validate the model.
We developed an innovative bayesian inversion approach to quantitatively estimate the uncertainties on the parameters avoiding the usual slow Monte Carlo approach. First we built lookup tables, and then searched the best fits and calculated a posteriori density probability functions. The results show that the model is able to reproduce the spectral behavior of water ice slabs, as well as the specular spot. In addition, the different parameters of the model are compatible with independent measurements.
△ Less
Submitted 15 July, 2015;
originally announced July 2015.
-
Radiative transfer model for contaminated rough slabs
Authors:
François Andrieu,
Sylvain Douté,
Frédéric Schmidt,
Bernard Schmitt
Abstract:
We present a semi-analytical model to simulate bidirectional reflectance distribution function (BRDF) spectra of a rough slab layer containing impurities. This model has been optimized for fast computation in order to analyze hyperspectral data. We designed it for planetary surfaces ices studies but it could be used for other purposes. It estimates the bidirectional reflectance of a rough slab of…
▽ More
We present a semi-analytical model to simulate bidirectional reflectance distribution function (BRDF) spectra of a rough slab layer containing impurities. This model has been optimized for fast computation in order to analyze hyperspectral data. We designed it for planetary surfaces ices studies but it could be used for other purposes. It estimates the bidirectional reflectance of a rough slab of material containing inclusions, overlaying an optically thick media (semi-infinite media or stratified media, for instance granular material). The inclusions are supposed to be close to spherical, and of any type of other material than the ice matrix. It can be any type of other ice, mineral or even bubbles, defined by their optical constants. We suppose a low roughness and we consider the geometrical optics conditions. This model is thus applicable for inclusions larger than the considered wavelength. The scattering on the inclusions is assumed to be isotropic. This model has a fast computation implementation and thus is suitable for high resolution hyperspectral data analysis.
△ Less
Submitted 10 June, 2015;
originally announced June 2015.
-
Time-dependent relaxation of strained silicon-on-insulator lines using a partially coherent x-ray nanobeam
Authors:
Francesca Mastropietro,
Joël Eymery,
Gerardina Carbone,
Sophie Baudot,
François Andrieu,
Vincent Favre-Nicolin
Abstract:
We report on the quantitative determination of the strain map in a strained Silicon-On-Insulator (sSOI) line with a 200x70 nm^2 cross-section. In order to study a single line as a function of time, we used an X-ray nanobeam with relaxed coherence properties as a compromise between beam size, coherence and intensity. We demonstrate how it is possible to reconstruct the line deformation at the nanos…
▽ More
We report on the quantitative determination of the strain map in a strained Silicon-On-Insulator (sSOI) line with a 200x70 nm^2 cross-section. In order to study a single line as a function of time, we used an X-ray nanobeam with relaxed coherence properties as a compromise between beam size, coherence and intensity. We demonstrate how it is possible to reconstruct the line deformation at the nanoscale, and follow its evolution as the line relaxes under the influence of the X-ray nanobeam.
△ Less
Submitted 14 June, 2013;
originally announced June 2013.
