Publications

Refine Results

(Filters Applied) Clear All

Learning to tutor from expert demonstrators via apprenticeship scheduling

February 4, 2017
  |
Conference Paper
Author:
Matthew C. Gombolay
…
Published in:
AAAI-17 Workshop on Human-Machine Collaborative Learning, 4 February 2017.
Topic:
machine learning
R&D area:
R&D group:

Summary

We have conducted a study investigating the use of automated tutors for educating players in the context of serious gaming (i.e., game designed as a professional training tool). Historically, researchers and practitioners have developed automated tutors through a process of manually codifying domain knowledge and translating that into a human-interpretable format. This process is laborious and leaves much to be desired. Instead, we seek to apply novel machine learning techniques to, first, learn a model from domain experts' demonstrations how to solve such problems, and, second, use this model to teach novices how to think like experts. In this work, we present a study comparing the performance of an automated and a traditional, manually-constructed tutor. To our knowledge, this is the first investigation using learning from demonstration techniques to learn from experts and use that knowledge to teach novices.
READ LESS

Summary

We have conducted a study investigating the use of automated tutors for educating players in the context of serious gaming (i.e., game designed as a professional training tool). Historically, researchers and practitioners have developed automated tutors through a process of manually codifying domain knowledge and translating that into a human-interpretable...

READ MORE
Learning to tutor from expert demonstrators via apprenticeship scheduling

WSR-88D chaff detection and characterization using an optimized hydrometeor classification algorithm

January 23, 2017
  |
Conference Paper
Author:
James M. Kurdzo
…
Published in:
18th Conf. on Aviation, Range, and Aerospace Meteorology, 23-26 January 2017.
Topic:
radar
R&D area:
R&D group:

Summary

Chaff presents multiple issues for aviation, air traffic controllers, and the FAA, including false weather identification and areas where flight paths may need to be altered. Chaff is a radar countermeasure commonly released from aircraft across the United States and is comprised of individual metallic strands designed to reflect certain wavelengths. Chaff returns tend to look similar to weather echoes in the reflectivity factor and radial velocity fields, and can appear as clutter, stratiform precipitation, or deep convection to the radar operator or radar algorithms. When polarimetric fields are taken into account, however, discrimination between weather and non-weather echoes has relatively high potential for success. In this work, the operational Hydrometeor Classification Algorithm (HCA) on the WSR-88D is modified to include a chaff class that can be used as input to a Chaff Detection Algorithm (CDA). This new class is designed using human-truthed chaff datasets for the collection and quantification of variable distributions, and the collected chaff cases are leveraged in the tuning of algorithm weights through the use of a metaheuristic optimization. A final CDA uses various image processing techniques to deliver a filtered output. A discussion regarding WSR-88D observations of chaff on a broad scale is provided, with particular attention given to observations of negative differential reflectivity during different stages of chaff fallout. Numerous cases are presented for analysis and characterization, both as an HCA class and as output from the filtered CDA.
READ LESS

Summary

Chaff presents multiple issues for aviation, air traffic controllers, and the FAA, including false weather identification and areas where flight paths may need to be altered. Chaff is a radar countermeasure commonly released from aircraft across the United States and is comprised of individual metallic strands designed to reflect certain...

READ MORE
WSR-88D chaff detection and characterization using an optimized hydrometeor classification algorithm

Learning by doing, High Performance Computing education in the MOOC era

January 17, 2017
  |
Journal Article
Author:
Julia Mullen
…
Published in:
J. Parallel Distrib. Comput., Vol. 105, July 2017, pp. 105-15.
Topic:
high performance computing
R&D area:
R&D group:

Summary

The High Performance Computing (HPC) community has spent decades developing tools that teach practitioners to harness the power of parallel and distributed computing. To create scalable and flexible educational experiences for practitioners in all phases of a career, we turn to Massively Open Online Courses (MOOCs). We detail the design of a unique self-paced online course that incorporates a focus on parallel solutions, personalization, and hands-on practice to familiarize student-users with their target system. Course material is presented through the lens of common HPC use cases and the strategies for parallelizing them. Using personalized paths, we teach researchers how to recognize the alignment between scientific applications and traditional HPC use cases, so they can focus on learning the parallelization strategies key to their workplace success. At the conclusion of their learning path, students should be capable of achieving performance gains on their HPC system.
READ LESS

Summary

The High Performance Computing (HPC) community has spent decades developing tools that teach practitioners to harness the power of parallel and distributed computing. To create scalable and flexible educational experiences for practitioners in all phases of a career, we turn to Massively Open Online Courses (MOOCs). We detail the design...

READ MORE
Learning by doing, High Performance Computing education in the MOOC era

Four-tap RF canceller evaluation for indoor in-band full-duplex wireless operation

January 15, 2017
  |
Conference Paper
Author:
Kenneth E. Kolodziej
…
Published in:
IEEE Radio and Wireless Symp., RWS 2017, 15-18 January 2017.
Topic:
wireless
R&D area:
R&D group:

Summary

Analog self-interference mitigation techniques are currently being investigated in a variety of operational settings for In-Band Full-Duplex (IBFD) systems. The significant multipath effects of realistic environments, such as inside buildings, can severely limit performance. The influence of different transceiver parameters on the effectiveness of a four-tap RF canceller using a tapped delay line architecture was characterized with a set of indoor measurements. The prototype canceller yielded up to 30 dB of cancellation over bandwidths ranging from 10 to 120 MHz centered at 2.45 GHz, and produced a combined analog system isolation that reached above 85 dB.
READ LESS

Summary

Analog self-interference mitigation techniques are currently being investigated in a variety of operational settings for In-Band Full-Duplex (IBFD) systems. The significant multipath effects of realistic environments, such as inside buildings, can severely limit performance. The influence of different transceiver parameters on the effectiveness of a four-tap RF canceller using a...

READ MORE
Four-tap RF canceller evaluation for indoor in-band full-duplex wireless operation

Spatially-resolved individual particle spectroscopy using photothermal modulation of Mie scattering

January 15, 2017
  |
Journal Article
Author:
Ryan M. Sullenberger
…
Published in:
Opt. Lett., Vol. 42, No. 2, 15 January 2017, pp. 203-6.
Topic:
optics
R&D area:
R&D group:

Summary

We report a photothermal modulation of Mie scattering (PMMS) method that enables concurrent spatial and spectral discrimination of individual micron-sized particles. This approach provides a direct measurement of the "fingerprint" infrared absorption spectrum with the spatial resolution of visible light. Trace quantities (tens of picograms) of material were deposited onto an infrared transparent substrate and simultaneously illuminated by a wavelength-tunable intensity-modulated quantum cascade pump laser and a continuous-wave 532 nm probe laser. Absorption of the pump laser by the particles results in direct modulation of the scatter field of the probe laser. The probe light scattered from the interrogated region is imaged onto a visible camera, enabling simultaneous probing of spatially-separated individual particles. By tuning the wavelength of the pump laser, the IR absorption spectrum is obtained. Using this approach, we measured the infrared absorption spectra of individual 3 um PMMA and silica spheres. Experimental PMMS signal amplitudes agree with modeling using an extended version of the Mie scattering theory for particles on substrates, enabling the prediction of the PMMS signal magnitude based on the material and substrate properties.
READ LESS

Summary

We report a photothermal modulation of Mie scattering (PMMS) method that enables concurrent spatial and spectral discrimination of individual micron-sized particles. This approach provides a direct measurement of the "fingerprint" infrared absorption spectrum with the spatial resolution of visible light. Trace quantities (tens of picograms) of material were deposited onto...

READ MORE
Spatially-resolved individual particle spectroscopy using photothermal modulation of Mie scattering

Interactive synthesis of code-level security rules

January 1, 2017
  |
Thesis
Author:
Leo St. Amour
Published in:
Thesis (M.S.)--Northeastern University, 2017.
Topic:
cyber security
R&D area:
R&D group:

Summary

Software engineers inadvertently introduce bugs into software during the development process and these bugs can potentially be exploited once the software is deployed. As the size and complexity of software systems increase, it is important that we are able to verify and validate not only that the software behaves as it is expected to, but also that it does not violate any security policies or properties. One of the approaches to reduce software vulnerabilities is to use a bug detection tool during the development process. Many bug detection techniques are limited by the burdensome and error prone process of manually writing a bug specification. Other techniques are able to learn specifications from examples, but are limited in the types of bugs that they are able to discover. This work presents a novel, general approach for synthesizing security rules for C code. The approach combines human knowledge with an interactive logic programming synthesis system to learn Datalog rules for various security properties. The approach has been successfully used to synthesize rules for three intraprocedural security properties: (1) out of bounds array accesses, (2) return value validation, and (3) double freed pointers. These rules have been evaluated on randomly generated C code and yield a 0% false positive rate and a 0%, 20%, and 0% false negative rate, respectively for each rule.
READ LESS

Summary

Software engineers inadvertently introduce bugs into software during the development process and these bugs can potentially be exploited once the software is deployed. As the size and complexity of software systems increase, it is important that we are able to verify and validate not only that the software behaves as...

READ MORE
Interactive synthesis of code-level security rules

Causal inference under network interference: a framework for experiments on social networks

January 1, 2017
  |
Thesis
Author:
Edward K. Kao
Published in:
Thesis (Ph.D.)--Harvard University, 2017.
Topic:
networking
R&D area:
R&D group:

Summary

No man is an island, as individuals interact and influence one another daily in our society. When social influence takes place in experiments on a population of interconnected individuals, the treatment on a unit may affect the outcomes of other units, a phenomenon known as interference. This thesis develops a causal framework and inference methodology for experiments where interference takes place on a network of influence (i.e. network interference). In this framework, the network potential outcomes serve as the key quantity and flexible building blocks for causal estimands that represent a variety of primary, peer, and total treatment effects. These causal estimands are estimated via principled Bayesian imputation of missing outcomes. The theory on the unconfoundedness assumptions leading to simplified imputation highlights the importance of including relevant network covariates in the potential outcome model. Additionally, experimental designs that result in balanced covariates and sizes across treatment exposure groups further improve the causal estimate, especially by mitigating potential outcome model mis-specification. The true potential outcome model is not typically known in real-world experiments, so the best practice is to account for interference and confounding network covariates through both balanced designs and model-based imputation. A full factorial simulated experiment is formulated to demonstrate this principle by comparing performance across different randomization schemes during the design phase and estimators during the analysis phase, under varying network topology and true potential outcome models. Overall, this thesis asserts that interference is not just a nuisance for analysis but rather an opportunity for quantifying and leveraging peer effects in real-world experiments.
READ LESS

Summary

No man is an island, as individuals interact and influence one another daily in our society. When social influence takes place in experiments on a population of interconnected individuals, the treatment on a unit may affect the outcomes of other units, a phenomenon known as interference. This thesis develops a...

READ MORE
Causal inference under network interference: a framework for experiments on social networks

Suppressing relaxation in superconducting qubits by quasiparticle pumping

December 23, 2016
  |
Journal Article
Author:
Simon Gustavsson
…
Published in:
Sci., Vol. 354, No. 6319, 23 December 2016, pp. 1573-77.
Topic:
quantum
R&D area:
R&D group:

Summary

Dynamical error suppression techniques are commonly used to improve coherence in quantum systems. They reduce dephasing errors by applying control pulses designed to reverse erroneous coherent evolution driven by environmental noise. However, such methods cannot correct for irreversible processes such as energy relaxation. We investigate a complementary, stochastic approach to reducting errors: instead of deterministically reversing the unwanted qubit evolution, we use control pulses to shape the noise environment dynamically. in the context of superconducting qubits, we implement a pumping sequence to reduce the number of unpaired electrons (quasiparticles) in close proximity to the device. A 70% reduction in the quasiparticle density reesults in a threefold enhancement in qubit relaxation times and a comparable reduction in coherence variablity.
READ LESS

Summary

Dynamical error suppression techniques are commonly used to improve coherence in quantum systems. They reduce dephasing errors by applying control pulses designed to reverse erroneous coherent evolution driven by environmental noise. However, such methods cannot correct for irreversible processes such as energy relaxation. We investigate a complementary, stochastic approach to...

READ MORE
Suppressing relaxation in superconducting qubits by quasiparticle pumping

Multi-lingual deep neural networks for language recognition

December 13, 2016
  |
Conference Paper
Author:
Luis R. Murphy Marcos
…
Published in:
SLT 2016, IEEE Spoken Language Technology Workshop, 13-16 December 2016.
Topic:
language recognition
R&D area:
R&D group:

Summary

Multi-lingual feature extraction using bottleneck layers in deep neural networks (BN-DNNs) has been proven to be an effective technique for low resource speech recognition and more recently for language recognition. In this work we investigate the impact on language recognition performance of the multi-lingual BN-DNN architecture and training configurations for the NIST 2011 and 2015 language recognition evaluations (LRE11 and LRE15). The best performing multi-lingual BN-DNN configuration yields relative performance gains of 50% on LRE11 and 40% on LRE15 compared to a standard MFCC/SDC baseline system and 17% on LRE11 and 7% on LRE15 relative to a single language BN-DNN system. Detailed performance analysis using data from all 24 Babel languages, Fisher Spanish and Switchboard English shows the impact of language selection and the amount of training data on overall BN-DNN performance.
READ LESS

Summary

Multi-lingual feature extraction using bottleneck layers in deep neural networks (BN-DNNs) has been proven to be an effective technique for low resource speech recognition and more recently for language recognition. In this work we investigate the impact on language recognition performance of the multi-lingual BN-DNN architecture and training configurations for...

READ MORE
Multi-lingual deep neural networks for language recognition

Approaches for language identification in mismatched environments

December 13, 2016
  |
Conference Paper
Author:
Shahan C. Nercessian
…
Published in:
SLT 2016, IEEE Spoken Language Technology Workshop, 13-16 December 2016.
Topic:
language recognition
R&D area:
R&D group:

Summary

In this paper, we consider the task of language identification in the context of mismatch conditions. Specifically, we address the issue of using unlabeled data in the domain of interest to improve the performance of a state-of-the-art system. The evaluation is performed on a 9-language set that includes data in both conversational telephone speech and narrowband broadcast speech. Multiple experiments are conducted to assess the performance of the system in this condition and a number of alternatives to ameliorate the drop in performance. The best system evaluated is based on deep neural network (DNN) bottleneck features using i-vectors utilizing a combination of all the approaches proposed in this work. The resulting system improved baseline DNN system performance by 30%.
READ LESS

Summary

In this paper, we consider the task of language identification in the context of mismatch conditions. Specifically, we address the issue of using unlabeled data in the domain of interest to improve the performance of a state-of-the-art system. The evaluation is performed on a 9-language set that includes data in...

READ MORE
Approaches for language identification in mismatched environments
401-410 of 2318