Publications
Speaker recognition using real vs synthetic parallel data for DNN channel compensation
September 8, 2016
Conference Paper
Published in:
INTERSPEECH 2016: 16th Annual Conf. of the Int. Speech Communication Assoc., 8-12 September 2016.
Topic:
R&D area:
Summary
Recent work has shown large performance gains using denoising DNNs for speech processing tasks under challenging acoustic conditions. However, training these DNNs requires large amounts of parallel multichannel speech data which can be impractical or expensive to collect. The effective use of synthetic parallel data as an alternative has been demonstrated for several speech technologies including automatic speech recognition and speaker recognition (SR). This paper demonstrates that denoising DNNs trained with real Mixer 2 multichannel data perform only slightly better than DNNs trained with synthetic multichannel data for microphone SR on Mixer 6. Large reductions in pooled error rates of 50% EER and 30% min DCF are achieved using DNNs trained on real Mixer 2 data. Nearly the same performance gains are achieved using synthetic data generated with a limited number of room impulse responses (RIRs) and noise sources derived from Mixer 2. Using RIRs from three publicly available sources used in the Kaldi ASpIRE recipe yields somewhat lower pooled gains of 34% EER and 25% min DCF. These results confirm the effective use of synthetic parallel data for DNN channel compensation even when the RIRs used for synthesizing the data are not particularly well matched to the task.
Summary
Recent work has shown large performance gains using denoising DNNs for speech processing tasks under challenging acoustic conditions. However, training these DNNs requires large amounts of parallel multichannel speech data which can be impractical or expensive to collect. The effective use of synthetic parallel data as an alternative has been...
READ MORE
Liquid crystal uncooled thermal imager development
August 28, 2016
Conference Paper
Published in:
SPIE, Vol. 9974, Infrared Sensors, Devices, and Applications VI, 28 August 2016.
Topic:
R&D area:
Summary
An uncooled thermal imager is being developed based on a liquid crystal (LC) transducer. Without any electrical connections, the LC transducer pixels change the long-wavelength infrared (LWIR) scene directly into a visible image as opposed to an electric signal in microbolometers. The objectives are to develop an imager technology scalable to large formats (tens of megapixels) while maintaining or improving the noise equivalent temperature difference (NETD) compared to microbolometers. The present work is demonstrating that the LCs have the required performance (sensitivity, dynamic range, speed, etc.) to enable a more flexible uncooled imager. Utilizing 200-mm wafers, a process has been developed and arrays have been fabricated using aligned LCs confined in 20-20-um cavities elevated on thermal legs. Detectors have been successfully fabricated on both silicon and fused silica wafers using less than 10 photolithographic mask steps. A breadboard camera system has been assembled to test the imagers. Various sensor configurations are described along with advantages and disadvantages of component arrangements.
Summary
An uncooled thermal imager is being developed based on a liquid crystal (LC) transducer. Without any electrical connections, the LC transducer pixels change the long-wavelength infrared (LWIR) scene directly into a visible image as opposed to an electric signal in microbolometers. The objectives are to develop an imager technology scalable...
READ MORE
The AFRL-MITLL WMT16 news-translation task systems
August 11, 2016
Conference Paper
Published in:
Proc. First Conf. on Machine Translation, Vol. 2, 11-12 August 2016, pp. 296-302.
Topic:
R&D area:
Summary
This paper describes the AFRL-MITLL statistical machine translation systems and the improvements that were developed during the WMT16 evaluation campaign. New techniques applied this year include Neural Machine Translation, a unique selection process for language modelling data, additional out-of-vocabulary transliteration techniques, and morphology generation.
Summary
This paper describes the AFRL-MITLL statistical machine translation systems and the improvements that were developed during the WMT16 evaluation campaign. New techniques applied this year include Neural Machine Translation, a unique selection process for language modelling data, additional out-of-vocabulary transliteration techniques, and morphology generation.
READ MORE
Matching community structure across online social networks
August 3, 2016
Journal Article
Published in:
arXiv, 3 August 2016.
Topic:
R&D area:
Summary
The discovery of community structure in networks is a problem of considerable interest in recent years. In online social networks, often times, users are simultaneously involved in multiple social media sites, some of which share common social relationships. It is of great interest to uncover a shared community structure across these networks. However, in reality, users typically identify themselves with different usernames across social media sites. This creates a great difficulty in detecting the community structure. In this paper, we explore several approaches for community detection across online social networks with limited knowledge of username alignment across the networks. We refer to the known alignment of usernames as seeds. We investigate strategies for seed selection and its impact on networks with a different fraction of overlapping vertices. The goal is to study the interplay between network topologies and seed selection strategies, and to understand how it affects the detected community structure. We also propose several measures to assess the performance of community detection and use them to measure the quality of the detected communities in both Twitter-Twitter networks and Twitter-Instagram networks.
Summary
The discovery of community structure in networks is a problem of considerable interest in recent years. In online social networks, often times, users are simultaneously involved in multiple social media sites, some of which share common social relationships. It is of great interest to uncover a shared community structure across...
READ MORE
Analytical models and methods for anomaly detection in dynamic, attributed graphs
July 31, 2016
Book Chapter
Published in:
Chapter 2, Computational Network Analysis with R: Applications in Biology, Medicine, and Chemistry, 2017, pp. 35-61.
Topic:
R&D area:
Summary
This chapter is devoted to anomaly detection in dynamic, attributed graphs. There has been a great deal of research on anomaly detection in graphs over the last decade, with a variety of methods proposed. This chapter discusses recent methods for anomaly detection in graphs,with a specific focus on detection within backgrounds based on random graph models. This sort of analysis can be applied for a variety of background models, which can incorporate topological dynamics and attributes of vertices and edges. The authors have developed a framework for anomalous subgraph detection in random background models, based on linear algebraic features of a graph. This includes an implementation in R that exploits structure in the random graph model for computationally tractable analysis of residuals. This chapter outlines this framework within the context of analyzing dynamic, attributed graphs. The remainder of this chapter is organized as follows. Section 2.2 defines the notation used within the chapter. Section 2.3 briefly describes a variety of perspectives and techniques for anomaly detection in graph-based data. Section 2.4 provides an overview of models for graph behavior that can be used as backgrounds for anomaly detection. Section 2.5 describes our framework for anomalous subgraph detection via spectral analysis of residuals, after the data are integrated over time. Section 2.6 discusses how the method described in Section 2.5 can be efficiently implemented in R using open source packages. Section 2.7 demonstrates the power of this technique in controlled simulation, considering the effects of both dynamics and attributes on detection performance. Section 2.8 gives a data analysis example within this context, using an evolving citation graph based on a commercially available document database of public scientific literature. Section 2.9 summarizes the chapter and discusses ongoing research in this area.
Summary
This chapter is devoted to anomaly detection in dynamic, attributed graphs. There has been a great deal of research on anomaly detection in graphs over the last decade, with a variety of methods proposed. This chapter discusses recent methods for anomaly detection in graphs,with a specific focus on detection within...
READ MORE
The TESS camera: modeling and measurements with deep depletion devices
July 29, 2016
Conference Paper
Published in:
Proc. SPIE 9904, Space Telescopes and Instrumentation 2016: Optical, Infrared, and Millimeter Wave, 29 July 2016, 99042C.
Topic:
R&D area:
Summary
The Transiting Exoplanet Survey Satellite, a NASA Explorer-class mission in development, will discover planets around nearby stars, most notably Earth-like planets with potential for follow up characterization. The all-sky survey requires a suite of four wide field-of-view cameras with sensitivity across a broad spectrum. Deep depletion CCDs with a silicon layer of 100 um thickness serve as the camera detectors, providing enhanced performance in the red wavelengths for sensitivity to cooler stars. The performance of the camera is critical for the mission objectives, with both the optical system and the CCD detectors contributing to the realized image quality. Expectations for image quality are studied using a combination of optical ray tracing in Zemax and simulations in Matlab to account for the interaction of the incoming photons with the 100 um silicon layer. The simulations include a probabilistic model to determine the depth of travel in the silicon before the photons are converted to photo-electrons, and a Monte Carlo approach to charge diffusion. The charge diffusion model varies with the remaining depth for the photo-electron to traverse and the strength of the intermediate electric field. The simulations are compared with laboratory measurements acquired by an engineering unit camera with the TESS optical design and deep depletion CCDs. In this paper we describe the performance simulations and the corresponding measurements taken with the engineering unit camera, and discuss where the models agree well in predicted trends and where there are differences compared to observations.
Summary
The Transiting Exoplanet Survey Satellite, a NASA Explorer-class mission in development, will discover planets around nearby stars, most notably Earth-like planets with potential for follow up characterization. The all-sky survey requires a suite of four wide field-of-view cameras with sensitivity across a broad spectrum. Deep depletion CCDs with a silicon...
READ MORE
Cross-domain entity resolution in social media
July 11, 2016
Conference Paper
Published in:
4th Int. Workshop on Natural Language Processing for Social Media, SocialNLP with IJCAI, 11 July 2016.
Summary
The challenge of associating entities across multiple domains is a key problem in social media understanding. Successful cross-domain entity resolution provides integration of information from multiple sites to create a complete picture of user and community activities, characteristics, and trends. In this work, we examine the problem of entity resolution across Twitter and Instagram using general techniques. Our methods fall into three categories: profile, content, and graph based. For the profile-based methods, we consider techniques based on approximate string matching. For content-based methods, we perform author identification. Finally, for graph-based methods, we apply novel cross-domain community detection methods and generate neighborhood-based features. The three categories of methods are applied to a large graph of users in Twitter and Instagram to understand challenges, determine performance, and understand fusion of multiple methods. Final results demonstrate an equal error rate less than 1%.
Summary
The challenge of associating entities across multiple domains is a key problem in social media understanding. Successful cross-domain entity resolution provides integration of information from multiple sites to create a complete picture of user and community activities, characteristics, and trends. In this work, we examine the problem of entity resolution...
READ MORE
Apprenticeship scheduling: learning to schedule from human experts
July 9, 2016
Conference Paper
Published in:
Proc. of the Int. Joint Conf. Artificial Intelligence (IJCAI), 9-15 July 2016.
Summary
Coordinating agents to complete a set of tasks with intercoupled temporal and resource constraints is computationally challenging, yet human domain experts can solve these difficult scheduling problems using paradigms learned through years of apprenticeship. A process for manually codifying this domain knowledge within a computational framework is necessary to scale beyond the "single expert, single-trainee" apprenticeship model. However, human domain experts often have difficulty describing their decision-making processes, causing the codification of this knowledge to become laborious. We propose a new approach for capturing domain-expert heuristics through a pairwise ranking formulation. Our approach is model-free and does not require enumerating or iterating through a large state-space. We empirically demonstrate that this approach accurately learns multifaceted heuristics on both a synthetic data set incorporating job-shop scheduling and vehicle routing problems and a real-world data set consisting of demonstrations of experts solving a weapon-to-target assignment problem.
Summary
Coordinating agents to complete a set of tasks with intercoupled temporal and resource constraints is computationally challenging, yet human domain experts can solve these difficult scheduling problems using paradigms learned through years of apprenticeship. A process for manually codifying this domain knowledge within a computational framework is necessary to scale...
READ MORE
Charting a security landscape in the clouds: data protection and collaboration in cloud storage
July 7, 2016
Technical Report
Published in:
MIT Lincoln Laboratory Report TR-1210
Topic:
R&D area:
R&D group:
Summary
This report surveys different approaches to securely storing and sharing data in the cloud based on traditional notions of security: confidentiality, integrity, and availability, with the main focus on confidentiality. An appendix discusses the related notion of how users can securely authenticate to cloud providers. We propose a metric for comparing secure storage approaches based on their residual vulnerabilities: attack surfaces against which an approach cannot protect. Our categorization therefore ranks approaches from the weakest (the most residual vulnerabilities) to the strongest (the fewest residual vulnerabilities). In addition to the security provided by each approach, we also consider their inherent costs and limitations. This report can therefore help an organization select a cloud data protection approach that satisfies their enterprise infrastructure, security specifications, and functionality requirements.
Summary
This report surveys different approaches to securely storing and sharing data in the cloud based on traditional notions of security: confidentiality, integrity, and availability, with the main focus on confidentiality. An appendix discusses the related notion of how users can securely authenticate to cloud providers. We propose a metric for...
READ MORE
The threat to weather radars by wireless technology
July 1, 2016
Journal Article
Published in:
Amer. Meteor. Soc., Vol. 97, No. 7, 1 July 2016, pp. 1159-67, doi: 10.1175/BAMS-D-15-00048.1.
Summary
Wireless technology, such as local area telecommunication networks and surveillance cameras, causes severe interference for weather radars, because they use the same operational radio frequencies. One or two disturbances can be removed from the radar image, but the number and power of the interfering wireless devices are growing all over the world, threatening that one day the radars could not be used at all. Some agencies have already changed or are considering changing frequency bands, but now even other bands are under threat. Use of equipment at radio frequencies is regulated by laws and international agreements. Technologies have been developed for peaceful co-existence. If wireless devices use these technologies to protect weather radars, their data transmission capabilities become limited, so it is tempting to violate the regulations. Hence, it is an important task for the worldwide weather community to involve themselves in the radio-frequency management process and work in close contact with their National Radio Authorities to ensure that meteorological interests be duly taken into account in any decision making process toward the future usage of wireless devices.
Summary
Wireless technology, such as local area telecommunication networks and surveillance cameras, causes severe interference for weather radars, because they use the same operational radio frequencies. One or two disturbances can be removed from the radar image, but the number and power of the interfering wireless devices are growing all over...
READ MORE