Publications
Combating Misinformation: HLT Highlights from MIT Lincoln Laboratory
March 17, 2021
Presentation
Published in:
Human Language Technology Conference (HLTCon), 16-18 March 2021.
R&D area:
Summary
Dr. Joseph Campbell shares several human language technologies highlights from MIT Lincoln Laboratory. These include key enabling technologies in combating misinformation to link personas, analyze content, and understand human networks. Developing operationally relevant technologies requires access to corresponding data with meaningful evaluations, as Dr. Douglas Reynolds presented in his keynote. As Dr. Danelle Shah discussed in her keynote, it’s crucial to develop these technologies to operate at deeper levels than the surface. Producing reliable information from the fusion of missing and inherently unreliable information channels is paramount. Furthermore, the dynamic misinformation environment and the coevolution of allied methods with adversarial methods represent additional challenges
Summary
Dr. Joseph Campbell shares several human language technologies highlights from MIT Lincoln Laboratory. These include key enabling technologies in combating misinformation to link personas, analyze content, and understand human networks. Developing operationally relevant technologies requires access to corresponding data with meaningful evaluations, as Dr. Douglas Reynolds presented in his keynote...
READ MORE
Combating Misinformation: What HLT Can (and Can't) Do When Words Don't Say What They Mean
March 17, 2021
Presentation
Published in:
Human Language Technology Conference (HLTCon), 16-18 March 2021.
R&D area:
Summary
Misinformation, disinformation, and “fake news” have been used as a means of influence for millennia, but the proliferation of the internet and social media in the 21st century has enabled nefarious campaigns to achieve unprecedented scale, speed, precision, and effectiveness. In the past few years, there has been significant recognition of the threats posed by malign influence operations to geopolitical relations, democratic institutions and processes, public health and safety, and more. At the same time, the digitization of communication offers tremendous opportunities for human language technologies (HLT) to observe, interpret, and understand this publicly available content. The ability to infer intent and impact, however, remains much more elusive.
Summary
Misinformation, disinformation, and “fake news” have been used as a means of influence for millennia, but the proliferation of the internet and social media in the 21st century has enabled nefarious campaigns to achieve unprecedented scale, speed, precision, and effectiveness. In the past few years, there has been significant recognition...
READ MORE
Detecting Parkinson's disease from wrist-worn accelerometry in the U.K. Biobank
March 14, 2021
Journal Article
Published in:
Sensors, Vol. 21, No. 6, 2021, Art. No. 2047.
Topic:
R&D area:
R&D group:
Summary
Parkinson's disease (PD) is a chronic movement disorder that produces a variety of characteristic movement abnormalities. The ubiquity of wrist-worn accelerometry suggests a possible sensor modality for early detection of PD symptoms and subsequent tracking of PD symptom severity. As an initial proof of concept for this technological approach, we analyzed the U.K. Biobank data set, consisting of one week of wrist-worn accelerometry from a population with a PD primary diagnosis and an age-matched healthy control population. Measures of movement dispersion were extracted from automatically segmented gait data, and measures of movement dimensionality were extracted from automatically segmented low-movement data. Using machine learning classifiers applied to one week of data, PD was detected with an area under the curve (AUC) of 0.69 on gait data, AUC = 0.84 on low-movement data, and AUC = 0.85 on a fusion of both activities. It was also found that classification accuracy steadily improved across the one-week data collection, suggesting that higher accuracy could be achievable from a longer data collection. These results suggest the viability of using a low-cost and easy-to-use activity sensor for detecting movement abnormalities due to PD and motivate further research on early PD detection and tracking of PD symptom severity.
Summary
Parkinson's disease (PD) is a chronic movement disorder that produces a variety of characteristic movement abnormalities. The ubiquity of wrist-worn accelerometry suggests a possible sensor modality for early detection of PD symptoms and subsequent tracking of PD symptom severity. As an initial proof of concept for this technological approach, we...
READ MORE
Multilayer microhydraulic actuators with speed and force configurations
March 11, 2021
Journal Article
Published in:
Microsyst. Nanoeng., Vol. 7, Art. No. 22, 2021.
Topic:
R&D area:
Summary
Electrostatic motors have traditionally required high voltage and provided low torque, leaving them with a vanishingly small portion of the motor application space. The lack of robust electrostatic motors is of particular concern in microsystems because inductive motors do not scale well to small dimensions. Often, microsystem designers have to choose from a host of imperfect actuation solutions, leading to high voltage requirements or low efficiency and thus straining the power budget of the entire system. In this work, we describe a scalable three-dimensional actuator technology that is based on the stacking of thin microhydraulic layers. This technology offers an actuation solution at 50 volts, with high force, high efficiency, fine stepping precision, layering, low abrasion, and resistance to pull-in instability. Actuator layers can also be stacked in different configurations trading off speed for force, and the actuator improves quadratically in power density when its internal dimensions are scaled-down.
Summary
Electrostatic motors have traditionally required high voltage and provided low torque, leaving them with a vanishingly small portion of the motor application space. The lack of robust electrostatic motors is of particular concern in microsystems because inductive motors do not scale well to small dimensions. Often, microsystem designers have to...
READ MORE
Multimodal representation learning via maximization of local mutual information [e-print]
March 7, 2021
Journal Article
Published in:
Intl. Conf. on Medical Image Computing and Computer Assisted Intervention, MICCAI, 27 September-1 October 2021.
Topic:
R&D area:
Summary
We propose and demonstrate a representation learning approach by maximizing the mutual information between local features of images and text. The goal of this approach is to learn useful image representations by taking advantage of the rich information contained in the free text that describes the findings in the image. Our method learns image and text encoders by encouraging the resulting representations to exhibit high local mutual information. We make use of recent advances in mutual information estimation with neural network discriminators. We argue that, typically, the sum of local mutual information is a lower bound on the global mutual information. Our experimental results in the downstream image classification tasks demonstrate the advantages of using local features for image-text representation learning.
Summary
We propose and demonstrate a representation learning approach by maximizing the mutual information between local features of images and text. The goal of this approach is to learn useful image representations by taking advantage of the rich information contained in the free text that describes the findings in the image...
READ MORE
Priority scheduling for multi-function apertures with hard- and soft-time constraints
March 6, 2021
Conference Paper
Published in:
2021 IEEE Aerospace Conf., 6-13 March 2021.
Summary
A multi-function aperture (MFA) is an antenna array that supports multiple RF signals for a diverse set of activities. An MFA may support multiple activities simultaneously if they are compatible, and platforms may utilize multiple MFAs to meet field-of-regard and frequency range requirements. Efficient MFA utilization requires a Resource Manager (RM) that routes signals to the correct MFA based on field-of-view and other requirements, and schedules MFA access to resolve conflicts based on request priority. An efficient RM scheduler time-interleaves requests from different activities as needed. Requested access events may be hard-time—that is, the event must be scheduled at a specified time or not at all, or soft-time, indicating it may be scheduled anytime in a specified window. Hard-time events include communications channels with assigned time slots, and soft-time events include asynchronous communications channels. This paper describes and evaluates an optimal algorithm to jointly schedule sequences of hard-time requests, maximizing the number of scheduled events while meeting priority requirements. An extension of this algorithm provides near-optimal schedules for sequences of soft-time or mixed hard- and soft-time events. Algorithms are evaluated by simulation, using two conflict models. The first is based on fixed signal paths that conflict if two paths share a common resource. The second model assumes the RM dynamically assigns resources. As implemented, these algorithms are too slow for real-time operation, and further work is required. They do provide insight into the MFA management problem, a useful metric for evaluating resource sharing and scheduling approaches, and may suggest efficient sub-optimal algorithms.
Summary
A multi-function aperture (MFA) is an antenna array that supports multiple RF signals for a diverse set of activities. An MFA may support multiple activities simultaneously if they are compatible, and platforms may utilize multiple MFAs to meet field-of-regard and frequency range requirements. Efficient MFA utilization requires a Resource Manager...
READ MORE
Using oculomotor features to predict changes in optic nerve sheath diameter and ImPACT scores from contact-sport athletes
March 4, 2021
Journal Article
Published in:
Front. Neurol., 4 March 2021, Art. No. 584684.
Topic:
R&D area:
Summary
There is mounting evidence linking the cumulative effects of repetitive head impacts to neuro-degenerative conditions. Robust clinical assessment tools to identify mild traumatic brain injuries are needed to assist with timely diagnosis for return-to-field decisions and appropriately guide rehabilitation. The focus of the present study is to investigate the potential for oculomotor features to complement existing diagnostic tools, such as measurements of Optic Nerve Sheath Diameter (ONSD) and Immediate Post-concussion Assessment and Cognitive Testing (ImPACT). Thirty-one high school American football and soccer athletes were tracked through the course of a sports season. Given the high risk of repetitive head impacts associated with both soccer and football, our hypotheses were that (1) ONSD and ImPACT scores would worsen through the season and (2) oculomotor features would effectively capture both neurophysiological changes reflected by ONSD and neuro-functional status assessed via ImPACT. Oculomotor features were used as input to Linear Mixed-Effects Regression models to predict ONSD and ImPACT scores as outcomes. Prediction accuracy was evaluated to identify explicit relationships between eye movements, ONSD, and ImPACT scores. Significant Pearson correlations were observed between predicted and actual outcomes for ONSD (Raw = 0.70; Normalized = 0.45) and for ImPACT (Raw = 0.86; Normalized = 0.71), demonstrating the capability of oculomotor features to capture neurological changes detected by both ONSD and ImPACT. The most predictive features were found to relate to motor control and visual-motor processing. In future work, oculomotor models, linking neural structures to oculomotor function, can be built to gain extended mechanistic insights into neurophysiological changes observed through seasons of participation in contact sports.
Summary
There is mounting evidence linking the cumulative effects of repetitive head impacts to neuro-degenerative conditions. Robust clinical assessment tools to identify mild traumatic brain injuries are needed to assist with timely diagnosis for return-to-field decisions and appropriately guide rehabilitation. The focus of the present study is to investigate the potential...
READ MORE
Speaker separation in realistic noise environments with applications to a cognitively-controlled hearing aid
March 4, 2021
Journal Article
Published in:
Neural Ntw., Vol. 140, 4 March 2021, pp. 136-47.
Summary
Future wearable technology may provide for enhanced communication in noisy environments and for the ability to pick out a single talker of interest in a crowded room simply by the listener shifting their attentional focus. Such a system relies on two components, speaker separation and decoding the listener's attention to acoustic streams in the environment. To address the former, we present a system for joint speaker separation and noise suppression, referred to as the Binaural Enhancement via Attention Masking Network (BEAMNET). The BEAMNET system is an end-to-end neural network architecture based on self-attention. Binaural input waveforms are mapped to a joint embedding space via a learned encoder, and separate multiplicative masking mechanisms are included for noise suppression and speaker separation. Pairs of output binaural waveforms are then synthesized using learned decoders, each capturing a separated speaker while maintaining spatial cues. A key contribution of BEAMNET is that the architecture contains a separation path, an enhancement path, and an autoencoder path. This paper proposes a novel loss function which simultaneously trains these paths, so that disabling the masking mechanisms during inference causes BEAMNET to reconstruct the input speech signals. This allows dynamic control of the level of suppression applied by BEAMNET via a minimum gain level, which is not possible in other state-of-the-art approaches to end-to-end speaker separation. This paper also proposes a perceptually-motivated waveform distance measure. Using objective speech quality metrics, the proposed system is demonstrated to perform well at separating two equal-energy talkers, even in high levels of background noise. Subjective testing shows an improvement in speech intelligibility across a range of noise levels, for signals with artificially added head-related transfer functions and background noise. Finally, when used as part of an auditory attention decoder (AAD) system using existing electroencephalogram (EEG) data, BEAMNET is found to maintain the decoding accuracy achieved with ideal speaker separation, even in severe acoustic conditions. These results suggest that this enhancement system is highly effective at decoding auditory attention in realistic noise environments, and could possibly lead to improved speech perception in a cognitively controlled hearing aid.
Summary
Future wearable technology may provide for enhanced communication in noisy environments and for the ability to pick out a single talker of interest in a crowded room simply by the listener shifting their attentional focus. Such a system relies on two components, speaker separation and decoding the listener's attention to...
READ MORE
Learning emergent discrete message communication for cooperative reinforcement learning
February 24, 2021
Conference Paper
Published in:
37th Conf. on Uncertainty in Artificial Intelligence, UAI 2021, early access, 26-30 July 2021.
Topic:
R&D area:
Summary
Communication is a important factor that enables agents work cooperatively in multi-agent reinforcement learning (MARL). Most previous work uses continuous message communication whose high representational capacity comes at the expense of interpretability. Allowing agents to learn their own discrete message communication protocol emerged from a variety of domains can increase the interpretability for human designers and other agents. This paper proposes a method to generate discrete messages analogous to human languages, and achieve communication by a broadcast-and-listen mechanism based on self-attention. We show that discrete message communication has performance comparable to continuous message communication but with much a much smaller vocabulary size. Furthermore, we propose an approach that allows humans to interactively send discrete messages to agents.
Summary
Communication is a important factor that enables agents work cooperatively in multi-agent reinforcement learning (MARL). Most previous work uses continuous message communication whose high representational capacity comes at the expense of interpretability. Allowing agents to learn their own discrete message communication protocol emerged from a variety of domains can increase...
READ MORE
More than a fair share: Network Data Remanence attacks against secret sharing-based schemes
February 23, 2021
Journal Article
Published in:
Network and Distributed Systems Security Symp., NDSS, 23-26 February 2021.
Topic:
R&D area:
R&D group:
Summary
With progress toward a practical quantum computer has come an increasingly rapid search for quantum-safe, secure communication schemes that do not rely on discrete logarithm or factorization problems. One such encryption scheme, Multi-path Switching with Secret Sharing (MSSS), combines secret sharing with multi-path switching to achieve security as long as the adversary does not have global observability of all paths and thus cannot capture enough shares to reconstruct messages. MSSS assumes that sending a share on a path is an atomic operation and all paths have the same delay. In this paper, we identify a side-channel vulnerability for MSSS, created by the fact that in real networks, sending a share is not an atomic operation as paths have multiple hops and different delays. This channel, referred to as Network Data Remanence (NDR), is present in all schemes like MSSS whose security relies on transfer atomicity and all paths having same delay. We demonstrate the presence of NDR in a physical testbed. We then identify two new attacks that aim to exploit the side channel, referred to as NDR Blind and NDR Planned, propose an analytical model to analyze the attacks, and demonstrate them using an implementation of MSSS based on the ONOS SDN controller. Finally, we present a countermeasure for the attacks and show its effectiveness in simulations and Mininet experiments.
Summary
With progress toward a practical quantum computer has come an increasingly rapid search for quantum-safe, secure communication schemes that do not rely on discrete logarithm or factorization problems. One such encryption scheme, Multi-path Switching with Secret Sharing (MSSS), combines secret sharing with multi-path switching to achieve security as long as...
READ MORE