Showing 1–26 of 26 results for author: Russell, L

LangProp: A code optimization framework using Large Language Models applied to driving

Authors: Shu Ishida, Gianluca Corrado, George Fedoseev, Hudson Yeo, Lloyd Russell, Jamie Shotton, João F. Henriques, Anthony Hu

Abstract: We propose LangProp, a framework for iteratively optimizing code generated by large language models (LLMs), in both supervised and reinforcement learning settings. While LLMs can generate sensible coding solutions zero-shot, they are often sub-optimal. Especially for code generation tasks, it is likely that the initial code will fail on certain edge cases. LangProp automatically evaluates the code… ▽ More We propose LangProp, a framework for iteratively optimizing code generated by large language models (LLMs), in both supervised and reinforcement learning settings. While LLMs can generate sensible coding solutions zero-shot, they are often sub-optimal. Especially for code generation tasks, it is likely that the initial code will fail on certain edge cases. LangProp automatically evaluates the code performance on a dataset of input-output pairs, catches any exceptions, and feeds the results back to the LLM in the training loop, so that the LLM can iteratively improve the code it generates. By adopting a metric- and data-driven training paradigm for this code optimization procedure, one could easily adapt findings from traditional machine learning techniques such as imitation learning, DAgger, and reinforcement learning. We show LangProp's applicability to general domains such as Sudoku and CartPole, as well as demonstrate the first proof of concept of automated code optimization for autonomous driving in CARLA. We show that LangProp can generate interpretable and transparent policies that can be verified and improved in a metric- and data-driven way. Our code is available at https://github.com/shuishida/LangProp. △ Less

Submitted 3 May, 2024; v1 submitted 18 January, 2024; originally announced January 2024.

GAIA-1: A Generative World Model for Autonomous Driving

Authors: Anthony Hu, Lloyd Russell, Hudson Yeo, Zak Murez, George Fedoseev, Alex Kendall, Jamie Shotton, Gianluca Corrado

Abstract: Autonomous driving promises transformative improvements to transportation, but building systems capable of safely navigating the unstructured complexity of real-world scenarios remains challenging. A critical problem lies in effectively predicting the various potential outcomes that may emerge in response to the vehicle's actions as the world evolves. To address this challenge, we introduce GAIA… ▽ More Autonomous driving promises transformative improvements to transportation, but building systems capable of safely navigating the unstructured complexity of real-world scenarios remains challenging. A critical problem lies in effectively predicting the various potential outcomes that may emerge in response to the vehicle's actions as the world evolves. To address this challenge, we introduce GAIA-1 ('Generative AI for Autonomy'), a generative world model that leverages video, text, and action inputs to generate realistic driving scenarios while offering fine-grained control over ego-vehicle behavior and scene features. Our approach casts world modeling as an unsupervised sequence modeling problem by map** the inputs to discrete tokens, and predicting the next token in the sequence. Emerging properties from our model include learning high-level structures and scene dynamics, contextual awareness, generalization, and understanding of geometry. The power of GAIA-1's learned representation that captures expectations of future events, combined with its ability to generate realistic samples, provides new possibilities for innovation in the field of autonomy, enabling enhanced and accelerated training of autonomous driving technology. △ Less

Submitted 29 September, 2023; originally announced September 2023.

Comments: Technical Report

Wide-Area Geolocalization with a Limited Field of View Camera in Challenging Urban Environments

Authors: Lena M. Downes, Ted J. Steiner, Rebecca L. Russell, Jonathan P. How

Abstract: Cross-view geolocalization, a supplement or replacement for GPS, localizes an agent within a search area by matching ground-view images to overhead images. Significant progress has been made assuming a panoramic ground camera. Panoramic cameras' high complexity and cost make non-panoramic cameras more widely applicable, but also more challenging since they yield less scene overlap between ground a… ▽ More Cross-view geolocalization, a supplement or replacement for GPS, localizes an agent within a search area by matching ground-view images to overhead images. Significant progress has been made assuming a panoramic ground camera. Panoramic cameras' high complexity and cost make non-panoramic cameras more widely applicable, but also more challenging since they yield less scene overlap between ground and overhead images. This paper presents Restricted FOV Wide-Area Geolocalization (ReWAG), a cross-view geolocalization approach that combines a neural network and particle filter to globally localize a mobile agent with only odometry and a non-panoramic camera. ReWAG creates pose-aware embeddings and provides a strategy to incorporate particle pose into the Siamese network, improving localization accuracy by a factor of 100 compared to a vision transformer baseline. This extended work also presents ReWAG*, which improves upon ReWAG's generalization ability in previously unseen environments. ReWAG* repeatedly converges accurately on a dataset of images we have collected in Boston with a 72 degree field of view (FOV) camera, a location and FOV that ReWAG* was not trained on. △ Less

Submitted 14 August, 2023; originally announced August 2023.

Comments: 10 pages, 16 figures. Extension of ICRA 2023 paper arXiv:2209.11854

Energetic electron precipitation driven by electromagnetic ion cyclotron waves from ELFIN's low altitude perspective

Authors: V. Angelopoulos, X. -J. Zhang, A. V. Artemyev, D. Mourenas, E. Tsai, C. Wilkins, A. Runov, J. Liu, D. L. Turner, W. Li, K. Khurana, R. E. Wirz, V. A. Sergeev, X. Meng, J. Wu, M. D. Hartinger, T. Raita, Y. Shen, X. An, X. Shi, M. F. Bashir, X. Shen, L. Gan, M. Qin, L. Capannolo , et al. (61 additional authors not shown)

Abstract: We review comprehensive observations of electromagnetic ion cyclotron (EMIC) wave-driven energetic electron precipitation using data from the energetic electron detector on the Electron Losses and Fields InvestigatioN (ELFIN) mission, two polar-orbiting low-altitude spinning CubeSats, measuring 50-5000 keV electrons with good pitch-angle and energy resolution. EMIC wave-driven precipitation exhibi… ▽ More We review comprehensive observations of electromagnetic ion cyclotron (EMIC) wave-driven energetic electron precipitation using data from the energetic electron detector on the Electron Losses and Fields InvestigatioN (ELFIN) mission, two polar-orbiting low-altitude spinning CubeSats, measuring 50-5000 keV electrons with good pitch-angle and energy resolution. EMIC wave-driven precipitation exhibits a distinct signature in energy-spectrograms of the precipitating-to-trapped flux ratio: peaks at 0.5 MeV which are abrupt (bursty) with significant substructure (occasionally down to sub-second timescale). Multiple ELFIN passes over the same MLT sector allow us to study the spatial and temporal evolution of the EMIC wave - electron interaction region. Using two years of ELFIN data, we assemble a statistical database of 50 events of strong EMIC wave-driven precipitation. Most reside at L=5-7 at dusk, while a smaller subset exists at L=8-12 at post-midnight. The energies of the peak-precipitation ratio and of the half-peak precipitation ratio (our proxy for the minimum resonance energy) exhibit an L-shell dependence in good agreement with theoretical estimates based on prior statistical observations of EMIC wave power spectra. The precipitation ratio's spectral shape for the most intense events has an exponential falloff away from the peak (i.e., on either side of 1.45 MeV). It too agrees well with quasi-linear diffusion theory based on prior statistics of wave spectra. Sub-MeV electron precipitation observed concurrently with strong EMIC wave-driven 1MeV precipitation has a spectral shape that is consistent with efficient pitch-angle scattering down to 200-300 keV by much less intense higher frequency EMIC waves. These results confirm the critical role of EMIC waves in driving relativistic electron losses. Nonlinear effects may abound and require further investigation. △ Less

Submitted 28 November, 2022; originally announced November 2022.

Wide-Area Geolocalization with a Limited Field of View Camera

Authors: Lena M. Downes, Ted J. Steiner, Rebecca L. Russell, Jonathan P. How

Abstract: Cross-view geolocalization, a supplement or replacement for GPS, localizes an agent within a search area by matching images taken from a ground-view camera to overhead images taken from satellites or aircraft. Although the viewpoint disparity between ground and overhead images makes cross-view geolocalization challenging, significant progress has been made assuming that the ground agent has access… ▽ More Cross-view geolocalization, a supplement or replacement for GPS, localizes an agent within a search area by matching images taken from a ground-view camera to overhead images taken from satellites or aircraft. Although the viewpoint disparity between ground and overhead images makes cross-view geolocalization challenging, significant progress has been made assuming that the ground agent has access to a panoramic camera. For example, our prior work (WAG) introduced changes in search area discretization, training loss, and particle filter weighting that enabled city-scale panoramic cross-view geolocalization. However, panoramic cameras are not widely used in existing robotic platforms due to their complexity and cost. Non-panoramic cross-view geolocalization is more applicable for robotics, but is also more challenging. This paper presents Restricted FOV Wide-Area Geolocalization (ReWAG), a cross-view geolocalization approach that generalizes WAG for use with standard, non-panoramic ground cameras by creating pose-aware embeddings and providing a strategy to incorporate particle pose into the Siamese network. ReWAG is a neural network and particle filter system that is able to globally localize a mobile agent in a GPS-denied environment with only odometry and a 90 degree FOV camera, achieving similar localization accuracy as what WAG achieved with a panoramic camera and improving localization accuracy by a factor of 100 compared to a baseline vision transformer (ViT) approach. A video highlight that demonstrates ReWAG's convergence on a test path of several dozen kilometers is available at https://youtu.be/U_OBQrt8qCE. △ Less

Submitted 18 May, 2023; v1 submitted 23 September, 2022; originally announced September 2022.

Comments: 7 pages, 10 figures. Accepted to ICRA 2023

Transverse charge density and the radius of the proton

Authors: Alexander V. Gramolin, Rebecca L. Russell

Abstract: A puzzling discrepancy exists between the values of the proton charge radius obtained using different experimental techniques: elastic electron-proton scattering and spectroscopy of electronic and muonic hydrogen. The proton radius is defined through the slope of the electric form factor, $G_E(Q^2)$, at zero four-momentum transfer, which is inaccessible in scattering experiments. We propose a nove… ▽ More A puzzling discrepancy exists between the values of the proton charge radius obtained using different experimental techniques: elastic electron-proton scattering and spectroscopy of electronic and muonic hydrogen. The proton radius is defined through the slope of the electric form factor, $G_E(Q^2)$, at zero four-momentum transfer, which is inaccessible in scattering experiments. We propose a novel method for extracting the proton radius from scattering data over a broad $Q^2$ range rather than attempting to directly determine the slope of $G_E$ at $Q^2 = 0$. This method relates the radius of the proton to its transverse charge density, which is the two-dimensional Fourier transform of the Dirac form factor, $F_1(Q^2)$. We apply our method to reanalyze the extensive data obtained by the A1 Collaboration [J. C. Bernauer et al., Phys. Rev. Lett. 105, 242001 (2010)] and extract a radius value, $r_E = 0.889(5)_{\text{stat}}(5)_{\text{syst}}(4)_{\text{model}}~\text{fm}$, that is consistent with the original result. We also provide new parametrizations for the Dirac and Pauli form factors and the transverse charge and magnetization densities of the proton. Our reanalysis shows that the proton radius discrepancy cannot be explained by issues with fitting and extrapolating the A1 data to $Q^2 = 0$. △ Less

Submitted 27 February, 2022; v1 submitted 25 February, 2021; originally announced February 2021.

Comments: Our Python analysis code is available at https://github.com/gramolin/radius/

Journal ref: Phys. Rev. D 105, 054004 (2022)

Optimising optical tweezers experiments for magnetic resonance sensing with nanodiamonds

Authors: L. W. Russell, E. C. Dossetor, A. A. Wood, D. A. Simpson, P. J. Reece

Abstract: In this article we explore the requirements for enabling high quality optically detected magnetic resonance (ODMR) spectroscopy in a conventional gradient force optical tweezers using nanodiamonds containing nitrogen-vacancy (NV$^{-}$) centres. We find that modulation of the infrared (1064 nm) trap** laser during spectroscopic measurements substantially improves the ODMR contrast compared with c… ▽ More In this article we explore the requirements for enabling high quality optically detected magnetic resonance (ODMR) spectroscopy in a conventional gradient force optical tweezers using nanodiamonds containing nitrogen-vacancy (NV$^{-}$) centres. We find that modulation of the infrared (1064 nm) trap** laser during spectroscopic measurements substantially improves the ODMR contrast compared with continuous wave trap**. The work is significant as it allows trap** and quantum sensing protocols to be performed in conditions where signal contrast is substantially reduced. We demonstrate the utility of the technique by resolving NV$^{-}$ spin projections within the ODMR spectrum. Manipulating the orientation of the nanodiamond via the trap** laser polarisation, we observe changes in spectral features. Theoretical modelling then allows us to infer the crystallographic orientation of the NV$^{-}$. This is an essential capability for future magnetic sensing applications of optically trapped nanodiamonds. △ Less

Submitted 21 January, 2021; originally announced January 2021.

An analysis pipeline for CHIME/FRB full-array baseband data

Authors: D. Michilli, K. W. Masui, R. Mckinven, D. Cubranic, M. Bruneault, C. Brar, C. Patel, P. J. Boyle, I. H. Stairs, A. Renard, K. Bandura, S. Berger, D. Breitman, T. Cassanelli, M. Dobbs, V. M. Kaspi, C. Leung, J. Mena-Parra, Z. Pleunis, L. Russell, P. Scholz, S. R. Siegel, S. P. Tendulkar, K. Vanderlinde

Abstract: The Canadian Hydrogen Intensity Map** Experiment (CHIME) has become a leading facility for detecting fast radio bursts (FRBs) through the CHIME/FRB backend. CHIME/FRB searches for fast transients in polarization-summed intensity data streams that have 24-kHz spectral and 1-ms temporal resolution. The intensity beams are pointed to pre-determined locations in the sky. A triggered baseband system… ▽ More The Canadian Hydrogen Intensity Map** Experiment (CHIME) has become a leading facility for detecting fast radio bursts (FRBs) through the CHIME/FRB backend. CHIME/FRB searches for fast transients in polarization-summed intensity data streams that have 24-kHz spectral and 1-ms temporal resolution. The intensity beams are pointed to pre-determined locations in the sky. A triggered baseband system records the coherent electric field measured by each antenna in the CHIME array at the time of FRB detections. Here we describe the analysis techniques and automated pipeline developed to process these full-array baseband data recordings. Whereas the real-time FRB detection pipeline has a localization limit of several arcminutes, offline analysis of baseband data yields source localizations with sub-arcminute precision, as characterized by using a sample of pulsars and one repeating FRB with known positions. The baseband pipeline also enables resolving temporal substructure on a micro-second scale and the study of polarization including detections of Faraday rotation. △ Less

Submitted 16 February, 2021; v1 submitted 13 October, 2020; originally announced October 2020.

The ELFIN Mission

Authors: V. Angelopoulos, E. Tsai, L. Bingley, C. Shaffer, D. L. Turner, A. Runov, W. Li, J. Liu, A. V. Artemyev, X. -J. Zhang, R. J. Strangeway, R. E. Wirz, Y. Y. Shprits, V. A. Sergeev, R. P. Caron, M. Chung, P. Cruce, W. Greer, E. Grimes, K. Hector, M. J. Lawson, D. Leneman, E. V. Masongsong, C. L. Russell, C. Wilkins , et al. (57 additional authors not shown)

Abstract: The Electron Loss and Fields Investigation with a Spatio-Temporal Ambiguity-Resolving option (ELFIN-STAR, or simply: ELFIN) mission comprises two identical 3-Unit (3U) CubeSats on a polar (~93deg inclination), nearly circular, low-Earth (~450 km altitude) orbit. Launched on September 15, 2018, ELFIN is expected to have a >2.5 year lifetime. Its primary science objective is to resolve the mechanism… ▽ More The Electron Loss and Fields Investigation with a Spatio-Temporal Ambiguity-Resolving option (ELFIN-STAR, or simply: ELFIN) mission comprises two identical 3-Unit (3U) CubeSats on a polar (~93deg inclination), nearly circular, low-Earth (~450 km altitude) orbit. Launched on September 15, 2018, ELFIN is expected to have a >2.5 year lifetime. Its primary science objective is to resolve the mechanism of storm-time relativistic electron precipitation, for which electromagnetic ion cyclotron (EMIC) waves are a prime candidate. From its ionospheric vantage point, ELFIN uses its unique pitch-angle-resolving capability to determine whether measured relativistic electron pitch-angle and energy spectra within the loss cone bear the characteristic signatures of scattering by EMIC waves or whether such scattering may be due to other processes. Pairing identical ELFIN satellites with slowly-variable along-track separation allows disambiguation of spatial and temporal evolution of the precipitation over minutes-to-tens-of-minutes timescales, faster than the orbit period of a single low-altitude satellite (~90min). Each satellite carries an energetic particle detector for electrons (EPDE) that measures 50keV to 5MeV electrons with deltaE/E<40% and a fluxgate magnetometer (FGM) on a ~72cm boom that measures magnetic field waves (e.g., EMIC waves) in the range from DC to 5Hz Nyquist (nominally) with <0.3nT/sqrt(Hz) noise at 1Hz. The spinning satellites (T_spin~3s) are equipped with magnetorquers that permit spin-up/down and reorientation maneuvers. The spin axis is placed normal to the orbit plane, allowing full pitch-angle resolution twice per spin. An energetic particle detector for ions (EPDI) measures 250keV-5MeV ions, addressing secondary science. Funded initially by CalSpace and the University Nanosat Program, ELFIN was selected for flight with joint support from NSF and NASA between 2014 and 2018. △ Less

Submitted 16 June, 2020; v1 submitted 13 June, 2020; originally announced June 2020.

Comments: Submitted to Space Science Reviews April 2020. 51 pages, 7 tables, 21 figures

Achieving Minimal Heat Conductivity by Ballistic Confinement in Phononic Metalattices

Authors: Weinan Chen, Disha Talreja, Devon Eichfeld, Pratibha Mahale, Nabila Nabi Nova, Hiu Y. Cheng, Jennifer L. Russell, Shih-Ying Yu, Nicolas Poilvert, Gerald Mahan, Suzanne E. Mohney, Vincent H. Crespi, Thomas E Mallouk, John V. Badding, Brian Foley, Venkatraman Gopalan, Ismaila Dabo

Abstract: Controlling the thermal conductivity of semiconductors is of practical interest in optimizing the performance of thermoelectric and phononic devices. The insertion of inclusions of nanometer size in a semiconductor is an effective means of achieving such control; it has been proposed that the thermal conductivity of silicon could be reduced to 1 W/m/K using this approach and that a minimum in the… ▽ More Controlling the thermal conductivity of semiconductors is of practical interest in optimizing the performance of thermoelectric and phononic devices. The insertion of inclusions of nanometer size in a semiconductor is an effective means of achieving such control; it has been proposed that the thermal conductivity of silicon could be reduced to 1 W/m/K using this approach and that a minimum in the heat conductivity would be reached for some optimal size of the inclusions. Yet the practical verification of this design rule has been limited. In this work, we address this question by studying the thermal properties of silicon metalattices that consist of a periodic distribution of spherical inclusions with radii from 7 to 30 nm, embedded into silicon. Experimental measurements confirm that the thermal conductivity of silicon metalattices is as low as 1 W/m/K for silica inclusions, and that this value can be further reduced to 0.16 W/m/K for silicon metalattices with empty pores. A detailed model of ballistic phonon transport suggests that this thermal conductivity is close to the lowest achievable by tuning the radius and spacing of the periodic inhomogeneities. This study is a significant step in elucidating the scaling laws that dictate ballistic heat transport at the nanoscale in silicon and other semiconductors. △ Less

Submitted 30 April, 2020; originally announced April 2020.

Journal ref: ACS Nano 2020, 14, 4, 4235-4243

Multivariate Uncertainty in Deep Learning

Authors: Rebecca L. Russell, Christopher Reale

Abstract: Deep learning has the potential to dramatically impact navigation and tracking state estimation problems critical to autonomous vehicles and robotics. Measurement uncertainties in state estimation systems based on Kalman and other Bayes filters are typically assumed to be a fixed covariance matrix. This assumption is risky, particularly for "black box" deep learning models, in which uncertainty ca… ▽ More Deep learning has the potential to dramatically impact navigation and tracking state estimation problems critical to autonomous vehicles and robotics. Measurement uncertainties in state estimation systems based on Kalman and other Bayes filters are typically assumed to be a fixed covariance matrix. This assumption is risky, particularly for "black box" deep learning models, in which uncertainty can vary dramatically and unexpectedly. Accurate quantification of multivariate uncertainty will allow for the full potential of deep learning to be used more safely and reliably in these applications. We show how to model multivariate uncertainty for regression problems with neural networks, incorporating both aleatoric and epistemic sources of heteroscedastic uncertainty. We train a deep uncertainty covariance matrix model in two ways: directly using a multivariate Gaussian density loss function, and indirectly using end-to-end training through a Kalman filter. We experimentally show in a visual tracking problem the large impact that accurate multivariate uncertainty quantification can have on Kalman filter performance for both in-domain and out-of-domain evaluation data. We additionally show in a challenging visual odometry problem how end-to-end filter training can allow uncertainty predictions to compensate for filter weaknesses. △ Less

Submitted 14 June, 2021; v1 submitted 30 October, 2019; originally announced October 2019.

Comments: To be published in IEEE Transactions on Neural Networks and Learning Systems

A 3D Probabilistic Deep Learning System for Detection and Diagnosis of Lung Cancer Using Low-Dose CT Scans

Authors: Onur Ozdemir, Rebecca L. Russell, Andrew A. Berlin

Abstract: We introduce a new computer aided detection and diagnosis system for lung cancer screening with low-dose CT scans that produces meaningful probability assessments. Our system is based entirely on 3D convolutional neural networks and achieves state-of-the-art performance for both lung nodule detection and malignancy classification tasks on the publicly available LUNA16 and Kaggle Data Science Bowl… ▽ More We introduce a new computer aided detection and diagnosis system for lung cancer screening with low-dose CT scans that produces meaningful probability assessments. Our system is based entirely on 3D convolutional neural networks and achieves state-of-the-art performance for both lung nodule detection and malignancy classification tasks on the publicly available LUNA16 and Kaggle Data Science Bowl challenges. While nodule detection systems are typically designed and optimized on their own, we find that it is important to consider the coupling between detection and diagnosis components. Exploiting this coupling allows us to develop an end-to-end system that has higher and more robust performance and eliminates the need for a nodule detection false positive reduction stage. Furthermore, we characterize model uncertainty in our deep learning systems, a first for lung CT analysis, and show that we can use this to provide well-calibrated classification probabilities for both nodule detection and patient malignancy diagnosis. These calibrated probabilities informed by model uncertainty can be used for subsequent risk-based decision making towards diagnostic interventions or disease treatments, as we demonstrate using a probability-based patient referral strategy to further improve our results. △ Less

Submitted 20 January, 2020; v1 submitted 8 February, 2019; originally announced February 2019.

Automated Vulnerability Detection in Source Code Using Deep Representation Learning

Authors: Rebecca L. Russell, Louis Kim, Lei H. Hamilton, Tomo Lazovich, Jacob A. Harer, Onur Ozdemir, Paul M. Ellingwood, Marc W. McConley

Abstract: Increasing numbers of software vulnerabilities are discovered every year whether they are reported publicly or discovered internally in proprietary code. These vulnerabilities can pose serious risk of exploit and result in system compromise, information leaks, or denial of service. We leveraged the wealth of C and C++ open-source code available to develop a large-scale function-level vulnerability… ▽ More Increasing numbers of software vulnerabilities are discovered every year whether they are reported publicly or discovered internally in proprietary code. These vulnerabilities can pose serious risk of exploit and result in system compromise, information leaks, or denial of service. We leveraged the wealth of C and C++ open-source code available to develop a large-scale function-level vulnerability detection system using machine learning. To supplement existing labeled vulnerability datasets, we compiled a vast dataset of millions of open-source functions and labeled it with carefully-selected findings from three different static analyzers that indicate potential exploits. The labeled dataset is available at: https://osf.io/d45bw/. Using these datasets, we developed a fast and scalable vulnerability detection tool based on deep feature representation learning that directly interprets lexed source code. We evaluated our tool on code from both real software packages and the NIST SATE IV benchmark dataset. Our results demonstrate that deep feature representation learning on source code is a promising approach for automated software vulnerability detection. △ Less

Submitted 27 November, 2018; v1 submitted 11 July, 2018; originally announced July 2018.

Learning to Repair Software Vulnerabilities with Generative Adversarial Networks

Authors: Jacob Harer, Onur Ozdemir, Tomo Lazovich, Christopher P. Reale, Rebecca L. Russell, Louis Y. Kim, Peter Chin

Abstract: Motivated by the problem of automated repair of software vulnerabilities, we propose an adversarial learning approach that maps from one discrete source domain to another target domain without requiring paired labeled examples or source and target domains to be bijections. We demonstrate that the proposed adversarial learning approach is an effective technique for repairing software vulnerabilitie… ▽ More Motivated by the problem of automated repair of software vulnerabilities, we propose an adversarial learning approach that maps from one discrete source domain to another target domain without requiring paired labeled examples or source and target domains to be bijections. We demonstrate that the proposed adversarial learning approach is an effective technique for repairing software vulnerabilities, performing close to seq2seq approaches that require labeled pairs. The proposed Generative Adversarial Network approach is application-agnostic in that it can be applied to other problems similar to code repair, such as grammar correction or sentiment translation. △ Less

Submitted 28 October, 2018; v1 submitted 18 May, 2018; originally announced May 2018.

Comments: Presented at 32nd Conference on Neural Information Processing Systems (nips 2018), Montreal Canada

Automated software vulnerability detection with machine learning

Authors: Jacob A. Harer, Louis Y. Kim, Rebecca L. Russell, Onur Ozdemir, Leonard R. Kosta, Akshay Rangamani, Lei H. Hamilton, Gabriel I. Centeno, Jonathan R. Key, Paul M. Ellingwood, Erik Antelman, Alan Mackay, Marc W. McConley, Jeffrey M. Opper, Peter Chin, Tomo Lazovich

Abstract: Thousands of security vulnerabilities are discovered in production software each year, either reported publicly to the Common Vulnerabilities and Exposures database or discovered internally in proprietary code. Vulnerabilities often manifest themselves in subtle ways that are not obvious to code reviewers or the developers themselves. With the wealth of open source code available for analysis, the… ▽ More Thousands of security vulnerabilities are discovered in production software each year, either reported publicly to the Common Vulnerabilities and Exposures database or discovered internally in proprietary code. Vulnerabilities often manifest themselves in subtle ways that are not obvious to code reviewers or the developers themselves. With the wealth of open source code available for analysis, there is an opportunity to learn the patterns of bugs that can lead to security vulnerabilities directly from data. In this paper, we present a data-driven approach to vulnerability detection using machine learning, specifically applied to C and C++ programs. We first compile a large dataset of hundreds of thousands of open-source functions labeled with the outputs of a static analyzer. We then compare methods applied directly to source code with methods applied to artifacts extracted from the build process, finding that source-based models perform better. We also compare the application of deep neural network models with more traditional models such as random forests and find the best performance comes from combining features learned by deep models with tree-based models. Ultimately, our highest performing model achieves an area under the precision-recall curve of 0.49 and an area under the ROC curve of 0.87. △ Less

Submitted 2 August, 2018; v1 submitted 14 February, 2018; originally announced March 2018.

Measurement and tricubic interpolation of the magnetic field for the OLYMPUS experiment

Authors: J. C. Bernauer, J. Diefenbach, G. Elbakian, G. Gavrilov, N. Goerrissen, D. K. Hasel, B. S. Henderson, Y. Holler, G. Karyan, J. Ludwig, H. Marukyan, Y. Naryshkin, C. O'Connor, R. L. Russell, A. Schmidt, U. Schneekloth, K. Suvorov, D. Veretennikov

Abstract: The OLYMPUS experiment used a 0.3 T toroidal magnetic spectrometer to measure the momenta of outgoing charged particles. In order to accurately determine particle trajectories, knowledge of the magnetic field was needed throughout the spectrometer volume. For that purpose, the magnetic field was measured at over 36,000 positions using a three-dimensional Hall probe actuated by a system of translat… ▽ More The OLYMPUS experiment used a 0.3 T toroidal magnetic spectrometer to measure the momenta of outgoing charged particles. In order to accurately determine particle trajectories, knowledge of the magnetic field was needed throughout the spectrometer volume. For that purpose, the magnetic field was measured at over 36,000 positions using a three-dimensional Hall probe actuated by a system of translation tables. We used these field data to fit a numerical magnetic field model, which could be employed to calculate the magnetic field at any point in the spectrometer volume. Calculations with this model were computationally intensive; for analysis applications where speed was crucial, we pre-computed the magnetic field and its derivatives on an evenly spaced grid so that the field could be interpolated between grid points. We developed a spline-based interpolation scheme suitable for SIMD implementations, with a memory layout chosen to minimize space and optimize the cache behavior to quickly calculate field values. This scheme requires only one-eighth of the memory needed to store necessary coefficients compared with a previous scheme [1]. This method was accurate for the vast majority of the spectrometer volume, though special fits and representations were needed to improve the accuracy close to the magnet coils and along the toroid axis. △ Less

Submitted 21 March, 2016; originally announced March 2016.

Comments: 7 pages, 8 figures

Journal ref: Nucl. Instrum. Meth. A 823, 9-14 (2016)

Simulations of Titan's paleoclimate

Authors: Juan M. Lora, Jonathan I. Lunine, Joellen L. Russell, Alexander G. Hayes

Abstract: We investigate the effects of varying Saturn's orbit on the atmospheric circulation and surface methane distribution of Titan. Using a new general circulation model of Titan's atmosphere, we simulate its climate under four characteristic configurations of orbital parameters that correspond to snapshots over the past 42 kyr, capturing the amplitude range of long-period cyclic variations in eccentri… ▽ More We investigate the effects of varying Saturn's orbit on the atmospheric circulation and surface methane distribution of Titan. Using a new general circulation model of Titan's atmosphere, we simulate its climate under four characteristic configurations of orbital parameters that correspond to snapshots over the past 42 kyr, capturing the amplitude range of long-period cyclic variations in eccentricity and longitude of perihelion. The model, which covers pressures from the surface to 0.5 mbar, reproduces the present-day temperature profile and tropospheric superrotation. In all four simulations, the atmosphere efficiently transports methane poleward, drying out the low- and mid-latitudes, indicating that these regions have been desert-like for at least tens of thousands of years. Though circulation patterns are not significantly different, the amount of surface methane that builds up over either pole strongly depends on the insolation distribution; in the present-day, methane builds up preferentially in the north, in agreement with observations, where summer is milder but longer. The same is true, to a lesser extent, for the configuration 14 kyr ago, while the south pole gains more methane in the case for 28 kyr ago, and the system is almost symmetric 42 kyr ago. This confirms the hypothesis that orbital forcing influences the distribution of surface liquids, and that the current observed asymmetry could have been partially or fully reversed in the past. The evolution of the orbital forcing implies that the surface reservoir is transported on timescales of $\sim$30 kyr, in which case the asymmetry reverses with a period of $\sim$125 kyr. Otherwise, the orbital forcing does not produce a net asymmetry over longer timescales, and is not a likely mechanism for generating the observed dichotomy. △ Less

Submitted 26 December, 2014; originally announced December 2014.

Journal ref: Icarus 243, 264-273 2014

GCM simulations of Titan's middle and lower atmosphere and comparison to observations

Authors: Juan M. Lora, Jonathan I. Lunine, Joellen L. Russell

Abstract: Simulation results are presented from a new general circulation model (GCM) of Titan, the Titan Atmospheric Model (TAM), which couples the Flexible Modeling System (FMS) spectral dynamical core to a suite of external/sub-grid-scale physics. These include a new non-gray radiative transfer module that takes advantage of recent data from Cassini-Huygens, large-scale condensation and quasi-equilibrium… ▽ More Simulation results are presented from a new general circulation model (GCM) of Titan, the Titan Atmospheric Model (TAM), which couples the Flexible Modeling System (FMS) spectral dynamical core to a suite of external/sub-grid-scale physics. These include a new non-gray radiative transfer module that takes advantage of recent data from Cassini-Huygens, large-scale condensation and quasi-equilibrium moist convection schemes, a surface model with "bucket" hydrology, and boundary layer turbulent diffusion. The model produces a realistic temperature structure from the surface to the lower mesosphere, including a stratopause, as well as satisfactory superrotation. The latter is shown to depend on the dynamical core's ability to build up angular momentum from surface torques. Simulated latitudinal temperature contrasts are adequate, compared to observations, and polar temperature anomalies agree with observations. In the lower atmosphere, the insolation distribution is shown to strongly impact turbulent fluxes, and surface heating is maximum at mid-latitudes. Surface liquids are unstable at mid- and low-latitudes, and quickly migrate poleward. The simulated humidity profile and distribution of surface temperatures, compared to observations, corroborate the prevalence of dry conditions at low latitudes. Polar cloud activity is well represented, though the observed mid-latitude clouds remain somewhat puzzling, and some formation alternatives are suggested. △ Less

Submitted 26 December, 2014; originally announced December 2014.

Comments: Accepted for publication in Icarus

An Integrated Experimental and Computational Investigation into the Dynamic Loads and Free-surface Wave-Field Perturbations Induced by Head-Sea Regular Waves on a 1/8.25 Scale-Model of the R/V ATHENA

Authors: Toby Ratcliffe, Lisa Minnick, Thomas T. O'Shea, Thomas Fu, Lauren Russell, Douglas G. Dommermuth

Abstract: A 1/8.25 scale-model of the U.S. Navy Research Vessel ATHENA was tested in regular head-sea waves to obtain data for validation of computational fluid dynamics (CFD) predictive tools. The experiments were performed in the David Taylor Model Basin at the Naval Surface Warfare Center (NSWC). With the model towed fixed in head-seas, horizontal and vertical loads on the model were obtained at two Frou… ▽ More A 1/8.25 scale-model of the U.S. Navy Research Vessel ATHENA was tested in regular head-sea waves to obtain data for validation of computational fluid dynamics (CFD) predictive tools. The experiments were performed in the David Taylor Model Basin at the Naval Surface Warfare Center (NSWC). With the model towed fixed in head-seas, horizontal and vertical loads on the model were obtained at two Froude numbers, $F_r=0.25$ and $F_r=0.43$. The model was run at two conditions of head-sea wavelengths corresponding to $λ=2L_o$ and $λ=1/2L_o$ with $H/λ=0.03$, where $L_o$ is the length of the model and $H=2 a$ is the wave height. The wave field perturbations induced by the head-sea waves were quantified from free-surface images generated by a laser light sheet. Predictions of the horizontal and vertical loads on the model in regular head sea waves were made with the Numerical Flow Analysis (NFA) code. Numerical predictions of the wave-field perturbations were compared with the experimental data and the correlation coefficients have been computed. △ Less

Submitted 7 October, 2014; originally announced October 2014.

Comments: 11 pages, 27th Symposium on Naval Hydrodynamics, Seoul, Korea, 5-10 October 2008

The OLYMPUS Experiment

Authors: R. Milner, D. K. Hasell, M. Kohl, U. Schneekloth, N. Akopov, R. Alarcon, V. A. Andreev, O. Ates, A. Avetisyan, D. Bayadilov, R. Beck, S. Belostotski, J. C. Bernauer, J. Bessuille, F. Brinker, B. Buck, J. R. Calarco, V. Carassiti, E. Cisbani, G. Ciullo, M. Contalbrigo, N. D'Ascenzo, R. De Leo, J. Diefenbach, T. W. Donnelly , et al. (48 additional authors not shown)

Abstract: The OLYMPUS experiment was designed to measure the ratio between the positron-proton and electron-proton elastic scattering cross sections, with the goal of determining the contribution of two-photon exchange to the elastic cross section. Two-photon exchange might resolve the discrepancy between measurements of the proton form factor ratio, $μ_p G^p_E/G^p_M$, made using polarization techniques and… ▽ More The OLYMPUS experiment was designed to measure the ratio between the positron-proton and electron-proton elastic scattering cross sections, with the goal of determining the contribution of two-photon exchange to the elastic cross section. Two-photon exchange might resolve the discrepancy between measurements of the proton form factor ratio, $μ_p G^p_E/G^p_M$, made using polarization techniques and those made in unpolarized experiments. OLYMPUS operated on the DORIS storage ring at DESY, alternating between 2.01~GeV electron and positron beams incident on an internal hydrogen gas target. The experiment used a toroidal magnetic spectrometer instrumented with drift chambers and time-of-flight detectors to measure rates for elastic scattering over the polar angular range of approximately $25^\circ$--$75^\circ$. Symmetric Møller/Bhabha calorimeters at $1.29^\circ$ and telescopes of GEM and MWPC detectors at $12^\circ$ served as luminosity monitors. A total luminosity of approximately 4.5~fb$^{-1}$ was collected over two running periods in 2012. This paper provides details on the accelerator, target, detectors, and operation of the experiment. △ Less

Submitted 5 December, 2013; originally announced December 2013.

Comments: 40 pages, 26 figures

Journal ref: Nuclear Inst. and Methods in Physics Research, A (2014) pp. 1-17

Investigation of a $^{85}$Rb Dark Magneto-Optical Trap using an Optical Nanofibre

Authors: L. Russell, R. Kumar, V. B. Tiwari, S. Nic Chormaic

Abstract: We report here measurements on a dark magneto-optical trap (DMOT) of $^{85}$Rb atoms using an optical nanofibre (ONF) with a waist of $\sim$~1 $μ$m. The DMOT is created using a doughnut-shaped repump beam along with a depump beam for efficient transfer of cold atoms from the bright hyperfine ground state ($F=3$) into the dark hyperfine ground state ($F=2$). The fluorescence from the cold $^{85}$Rb… ▽ More We report here measurements on a dark magneto-optical trap (DMOT) of $^{85}$Rb atoms using an optical nanofibre (ONF) with a waist of $\sim$~1 $μ$m. The DMOT is created using a doughnut-shaped repump beam along with a depump beam for efficient transfer of cold atoms from the bright hyperfine ground state ($F=3$) into the dark hyperfine ground state ($F=2$). The fluorescence from the cold $^{85}$Rb atoms of the DMOT is detected by coupling it into the fibre-guided modes of the ONF. The measured fractional population of cold atoms in the bright hyperfine ground state ($p$) is as low as $\sim$0.04. The dependence of loading rate of DMOT on cooling laser intensity is investigated and also compared with the loading rate of a bright-MOT (BMOT). This work lays the foundation for the use of an ONF for probing of a small number of atoms in an optically-dense cold atomic cloud. △ Less

Submitted 1 November, 2013; v1 submitted 24 June, 2013; originally announced June 2013.

Comments: 8 pages, 8 figures, minor changes made

MSC Class: 81V45

Spectroscopy, Manipulation and Trap** of Neutral Atoms, Molecules, and Other Particles using Optical Nanofibers: A Review

Authors: Michael J. Morrissey, Kieran Deasy, Mary Frawley, Ravi Kumar, Eugen Prel, Laura Russell, Viet Giang Truong, Síle Nic Chormaic

Abstract: The use of tapered optical fibers, i.e., optical nanofibers, for spectroscopy and the detection of small numbers of particles, such as neutral atoms or molecules, has been gaining ground in recent years. In this review, we briefly introduce the optical nanofiber, its fabrication and optical mode propagation within. We discuss recent progress on the integration of optical nanofibers into laser-cool… ▽ More The use of tapered optical fibers, i.e., optical nanofibers, for spectroscopy and the detection of small numbers of particles, such as neutral atoms or molecules, has been gaining ground in recent years. In this review, we briefly introduce the optical nanofiber, its fabrication and optical mode propagation within. We discuss recent progress on the integration of optical nanofibers into laser-cooled atom and vapor systems, paying particular attention to spectroscopy, cold atom cloud characterization and optical trap** schemes. Next, a natural extension on this work to molecules will be introduced. Finally, we consider several alternatives to optical nanofibers that display some advantages for particular applications. △ Less

Submitted 24 June, 2013; originally announced June 2013.

Comments: 28 pages, 12 figures, review article

MSC Class: 81V45

Measurements on release-recapture of cold Rb-85 atoms using an optical nanofibre in a magneto-optical trap

Authors: L. Russell, R. Kumar, V. B. Tiwari, S. Nic Chormaic

Abstract: We have performed release-recapture temperature measurements of laser-cooled Rb-85 atoms using an optical nanofibre (ONF) in a magneto-optical trap (MOT). The effects of changing the cooling laser light-shift parameter on the temperature of the cold atoms and spring constant of the trap are studied. By varying the cold atom number density near the ONF, the onset of the multiple scattering regime i… ▽ More We have performed release-recapture temperature measurements of laser-cooled Rb-85 atoms using an optical nanofibre (ONF) in a magneto-optical trap (MOT). The effects of changing the cooling laser light-shift parameter on the temperature of the cold atoms and spring constant of the trap are studied. By varying the cold atom number density near the ONF, the onset of the multiple scattering regime is observed without the need for an estimation of the atom cloud size. Moreover, this sensitive ONF assisted release-recapture technique is easily able to detect any optical misalignment of the cooling laser beams in the MOT. △ Less

Submitted 1 August, 2013; v1 submitted 29 March, 2013; originally announced March 2013.

Comments: 17 pages, 7 figures

Observation of double resistance anomalies and excessive resistance in mesoscopic superconducting Au$_{0.7}$In$_{0.3}$ rings with phase separation

Authors: H. Wang, M. M. Rosario, H. L. Russell, Y. Liu

Abstract: We have measured mesoscopic superconducting Au$_{0.7}$In$_{0.3}$ rings prepared by e-beam lithography and sequential deposition of Au and In at room temperature followed by a standard lift-off procedure. In samples showing no Little-Parks resistance oscillations, highly unusual double resistance anomalies, two resistance peaks found near the onset of superconductivity, were observed. Although re… ▽ More We have measured mesoscopic superconducting Au$_{0.7}$In$_{0.3}$ rings prepared by e-beam lithography and sequential deposition of Au and In at room temperature followed by a standard lift-off procedure. In samples showing no Little-Parks resistance oscillations, highly unusual double resistance anomalies, two resistance peaks found near the onset of superconductivity, were observed. Although resistance anomaly featuring a single resistance peak has been seen in various mesoscopic superconducting samples, double resistance anomalies have never been observed previously. The dynamical resistance measurements suggest that there are two critical currents in these samples. In addition, the two resistance peaks were found to be suppressed at different magnetic fields. We attribute the observed double resistance anomalies to an underlying phase separation in which In-rich grains of intermetallic compound of AuIn precipitate in a uniform In-dilute matrix of Au$_{0.9}$In$_{0.1}$. The intrinsic superconducting transition temperature of the In-rich grains is substantially higher than that of the In-dilute matrix. The suppression of the conventional Little-Parks resistance oscillation is explained in the same picture by taking into consideration a strong variation in the $T_c$ of the In-rich grains. We also report the observation of an unusual magnetic-field-induced metallic state with its resistance higher than the normal-state resistance, referred to here as excessive resistance, and an h/2e resistance oscillation with the amplitude of oscillation depends extremely weakly on temperature. △ Less

Submitted 19 October, 2005; v1 submitted 21 October, 2004; originally announced October 2004.

Comments: 6 pages, 7 figures, changed content

Ideological and Policy Origins of the Internet, 1957-1969

Authors: Andrew L. Russell

Abstract: This paper examines the ideological and policy consensus that shaped computing research funded by the Information Processing Techniques Office (IPTO) within the Department of Defense's Advanced Research Projects Agency (ARPA). This historical case study of the period between Sputnik and the creation of the ARPANET shows how military, scientific, and academic values shaped the institutions and re… ▽ More This paper examines the ideological and policy consensus that shaped computing research funded by the Information Processing Techniques Office (IPTO) within the Department of Defense's Advanced Research Projects Agency (ARPA). This historical case study of the period between Sputnik and the creation of the ARPANET shows how military, scientific, and academic values shaped the institutions and relations of a foundational period in the creation of the Internet. The paper probes three areas: the ideology of the science policy consensus, the institutional philosophy of IPTO under J. C. R. Licklider, and the ways that this consensus and philosophy shaped IPTO research in the period leading to the creation of the ARPANET. By examining the intellectual, cultural, and institutional details of the consensus that governed IPTO research between 1957 and 1969, we can understand the ways that these values defined the range of possibilities for network computing. The influence of the social values expressed by these actors was decisive: that government had an obligation to support a broad base of scientific research to promote both the public good and the national defense; that IPTO-sponsored computing research would accomplish both military and scientific objectives; and that IPTO could leverage its power within this consensus to create a network to share resources and unite researchers over geographical distance. A greater awareness of the ways that "consensus" worked in this period -- the "pre-history" of the Internet -- provides a richer context for evaluating the unique features of the Internet, such as its open architecture, collegial culture, and standards-based governance. △ Less

Submitted 24 October, 2001; v1 submitted 24 September, 2001; originally announced September 2001.

Comments: 29th TPRC Conference, 2001

Report number: TPRC-2001-087 ACM Class: K.4.m Miscellaneous

RORF - a Radio Optical Reference Frame

Authors: N. Zacharias, A. L. Fey, J. L. Russell, K. J. Johnston

Abstract: The radio observations are based on dual frequency Mark--III VLBI data from 1979 until the end of 1993. Currently 436 sources define the celestial inertial frame. A comparison to the JPL94R01 catalog is made. Position vs. time plots of selected sources show significant apparent non linear proper motions. The status of the project and recent results from optical observations are presented. The radio observations are based on dual frequency Mark--III VLBI data from 1979 until the end of 1993. Currently 436 sources define the celestial inertial frame. A comparison to the JPL94R01 catalog is made. Position vs. time plots of selected sources show significant apparent non linear proper motions. The status of the project and recent results from optical observations are presented. △ Less

Submitted 29 November, 1994; originally announced November 1994.

Comments: 1 page (LaTeX), IAU Symp.166, The radio catalog is available at ftp://maia.usno.navy.mil/rorf , and complete poster paper is available as hardcopy from [email protected]