Publications
Beyond expertise and roles: a framework to characterize the stakeholders of interpretable machine learning and their needs
January 24, 2021
Conference Paper
Published in:
Proc. Conf. on Human Factors in Computing Systems, 8-13 May 2021, article no. 74.
Summary
To ensure accountability and mitigate harm, it is critical that diverse stakeholders can interrogate black-box automated systems and find information that is understandable, relevant, and useful to them. In this paper, we eschew prior expertise- and role-based categorizations of interpretability stakeholders in favor of a more granular framework that decouples stakeholders' knowledge from their interpretability needs. We characterize stakeholders by their formal, instrumental, and personal knowledge and how it manifests in the contexts of machine learning, the data domain, and the general milieu. We additionally distill a hierarchical typology of stakeholder needs that distinguishes higher-level domain goals from lower-level interpretability tasks. In assessing the descriptive, evaluative, and generative powers of our framework, we find our more nuanced treatment of stakeholders reveals gaps and opportunities in the interpretability literature, adds precision to the design and comparison of user studies, and facilitates a more reflexive approach to conducting this research.
Summary
To ensure accountability and mitigate harm, it is critical that diverse stakeholders can interrogate black-box automated systems and find information that is understandable, relevant, and useful to them. In this paper, we eschew prior expertise- and role-based categorizations of interpretability stakeholders in favor of a more granular framework that decouples...
READ MORE
Seasonal Inhomogeneous Nonconsecutive Arrival Process Search and Evaluation
January 15, 2021
Conference Paper
Published in:
25th International Conference on Pattern Recognition [submitted]
Topic:
R&D area:
Summary
Time series often exhibit seasonal patterns, and identification of these patterns is essential to understanding thedata and predicting future behavior. Most methods train onlarge datasets and can fail to predict far past the training data. This limitation becomes more pronounced when data is sparse. This paper presents a method to fit a model to seasonal time series data that maintains predictive power when data is limited. This method, called SINAPSE, combines statistical model fitting with an information criteria to search for disjoint, andpossibly nonconsecutive, regimes underlying the data, allowing for a sparse representation resistant to overfitting.
Summary
Time series often exhibit seasonal patterns, and identification of these patterns is essential to understanding thedata and predicting future behavior. Most methods train onlarge datasets and can fail to predict far past the training data. This limitation becomes more pronounced when data is sparse. This paper presents a method to...
READ MORE
Automatic detection of influential actors in disinformation networks
January 7, 2021
Journal Article
Published in:
Proc. Natl. Acad. Sci., Vol. 118, No. 4, January 2021, e2011216118.
Topic:
R&D area:
Summary
The weaponization of digital communications and social media to conduct disinformation campaigns at immense scale, speed, and reach presents new challenges to identify and counter hostile influence operations (IO). This paper presents an end-to-end framework to automate detection of disinformation narratives, networks, and influential actors. The framework integrates natural language processing, machine learning, graph analytics, and a novel network causal inference approach to quantify the impact of individual actors in spreading IO narratives. We demonstrate its capability on real-world hostile IO campaigns with Twitter datasets collected during the 2017 French presidential elections, and known IO accounts disclosed by Twitter. Our system detects IO accounts with 96% precision, 79% recall, and 96% area-under-the-PR-curve, maps out salient network communities, and discovers high-impact accounts that escape the lens of traditional impact statistics based on activity counts and network centrality. Results are corroborated with independent sources of known IO accounts from U.S. Congressional reports, investigative journalism, and IO datasets provided by Twitter.
Summary
The weaponization of digital communications and social media to conduct disinformation campaigns at immense scale, speed, and reach presents new challenges to identify and counter hostile influence operations (IO). This paper presents an end-to-end framework to automate detection of disinformation narratives, networks, and influential actors. The framework integrates natural language...
READ MORE
The Speech Enhancement via Attention Masking Network (SEAMNET): an end-to-end system for joint suppression of noise and reverberation [early access]
January 1, 2021
Journal Article
Published in:
IEEE/ACM Trans. on Audio, Speech, and Language Processing, Vol. 29, 2021, pp. 515-26.
Topic:
R&D area:
Summary
This paper proposes the Speech Enhancement via Attention Masking Network (SEAMNET), a neural network-based end-to-end single-channel speech enhancement system designed for joint suppression of noise and reverberation. It formalizes an end-to-end network architecture, referred to as b-Net, which accomplishes noise suppression through attention masking in a learned embedding space. A key contribution of SEAMNET is that the b-Net architecture contains both an enhancement and an autoencoder path. This paper proposes a novel loss function which simultaneously trains both the enhancement and the autoencoder paths, so that disabling the masking mechanism during inference causes SEAMNET to reconstruct the input speech signal. This allows dynamic control of the level of suppression applied by SEAMNET via a minimum gain level, which is not possible in other state-of-the-art approaches to end-to-end speech enhancement. This paper also proposes a perceptually-motivated waveform distance measure. In addition to the b-Net architecture, this paper proposes a novel method for designing target waveforms for network training, so that joint suppression of additive noise and reverberation can be performed by an end-to-end enhancement system, which has not been previously possible. Experimental results show the SEAMNET system to outperform a variety of state-of-the-art baselines systems, both in terms of objective speech quality measures and subjective listening tests. Finally, this paper draws parallels between SEAMNET and conventional statistical model-based enhancement approaches, offering interpretability of many network components.
Summary
This paper proposes the Speech Enhancement via Attention Masking Network (SEAMNET), a neural network-based end-to-end single-channel speech enhancement system designed for joint suppression of noise and reverberation. It formalizes an end-to-end network architecture, referred to as b-Net, which accomplishes noise suppression through attention masking in a learned embedding space. A...
READ MORE
Information Aware max-norm Dirichlet networks for predictive uncertainty estimation
January 1, 2021
Journal Article
Published in:
Neural Netw., Vol. 135, 2021, pp. 105–114.
Topic:
R&D group:
Summary
Precise estimation of uncertainty in predictions for AI systems is a critical factor in ensuring trust and safety. Deep neural networks trained with a conventional method are prone to over-confident predictions. In contrast to Bayesian neural networks that learn approximate distributions on weights to infer prediction confidence, we propose a novel method, Information Aware Dirichlet networks, that learn an explicit Dirichlet prior distribution on predictive distributions by minimizing a bound on the expected max norm of the prediction error and penalizing information associated with incorrect outcomes. Properties of the new cost function are derived to indicate how improved uncertainty estimation is achieved. Experiments using real datasets show that our technique outperforms, by a large margin, state-of-the-art neural networks for estimating within-distribution and out-of-distribution uncertainty, and detecting adversarial examples.
Summary
Precise estimation of uncertainty in predictions for AI systems is a critical factor in ensuring trust and safety. Deep neural networks trained with a conventional method are prone to over-confident predictions. In contrast to Bayesian neural networks that learn approximate distributions on weights to infer prediction confidence, we propose a...
READ MORE
Ablation analysis to select wearable sensors for classifying standing, walking, and running
December 30, 2020
Journal Article
Published in:
Sensors, Vol. 21, No. 1, 30 December 2020, 194.
Summary
The field of human activity recognition (HAR) often utilizes wearable sensors and machine learning techniques in order to identify the actions of the subject. This paper considers the activity recognition of walking and running while using a support vector machine (SVM) that was trained on principal components derived from wearable sensor data. An ablation analysis is performed in order to select the subset of sensors that yield the highest classification accuracy. The paper also compares principal components across trials to inform the similarity of the trials. Five subjects were instructed to perform standing, walking, running, and sprinting on a self-paced treadmill, and the data were recorded while using surface electromyography sensors (sEMGs), inertial measurement units (IMUs), and force plates. When all of the sensors were included, the SVM had over 90% classification accuracy using only the first three principal components of the data with the classes of stand, walk, and run/sprint (combined run and sprint class). It was found that sensors that were placed only on the lower leg produce higher accuracies than sensors placed on the upper leg. There was a small decrease in accuracy when the force plates are ablated, but the difference may not be operationally relevant. Using only accelerometers without sEMGs was shown to decrease the accuracy of the SVM.
Summary
The field of human activity recognition (HAR) often utilizes wearable sensors and machine learning techniques in order to identify the actions of the subject. This paper considers the activity recognition of walking and running while using a support vector machine (SVM) that was trained on principal components derived from wearable...
READ MORE
NASA Airspace Integration Detect and Avoid Phase 2: Safety Risk Management Simulation Plan
December 18, 2020
Project Report
Published in:
MIT Lincoln Laboratory Report
Topic:
R&D area:
R&D group:
Summary
RTCA has been developing Minimum Operational Performance Standards (MOPS) for Detect and Avoid (DAA) and Command and Control (C2) systems as part of Special Committee – 228 (SC-228). The Phase 1 MOPS were published in 2017 and a Phase 2 effort to revise and extend the Phase 1 MOPS is ongoing. In order for the MOPS to be fully utilized, they must be evaluated by the FAA employing the FAA's Safety Risk Management (SRM) process. In order to support the SRM process, there is a need for simulation data focused on the safety of DAA encounters. This analysis focuses on gathering information to validate the use of SC-228 MOPS compliant DAA and C2 systems to enable routine UAS operations in the National Airspace System (NAS) without a chase aircraft or visual observers. The scope of this effort aligns with the SC-228 Terms of Reference (TOR) that can be generally characterized as UAS flying IFR and receiving ATC separation services. This analysis evaluates the system safety in mixed classes B, C, D, E, and G airspaces, and includes IFR, VFR, Cooperative, and Non-Cooperative aircraft. This analysis plan describes the four analysis tasks (Section 2) and the simulation plan (Section 3) that will be executed to accomplish these tasks. It is expected that this analysis plan will be extended to include the analysis results and become the final deliverable.
Summary
RTCA has been developing Minimum Operational Performance Standards (MOPS) for Detect and Avoid (DAA) and Command and Control (C2) systems as part of Special Committee – 228 (SC-228). The Phase 1 MOPS were published in 2017 and a Phase 2 effort to revise and extend the Phase 1 MOPS is...
READ MORE
Operation of an optical atomic clock with a Brillouin laser subsystem
December 9, 2020
Journal Article
Published in:
Nature, Vol. 588, No. 7837, 2020, pp. 244-9.
Topic:
R&D area:
R&D group:
Summary
Microwave atomic clocks have traditionally served as the 'gold standard' for precision measurements of time and frequency. However, over the past decade, optical atomic clocks have surpassed the precision of their microwave counterparts by two orders of magnitude or more. Extant optical clocks occupy volumes of more than one cubic metre, and it is a substantial challenge to enable these clocks to operate in field environments, which requires the ruggedization and miniaturization of the atomic reference and clock laser along with their supporting lasers and electronics. In terms of the clock laser, prior laboratory demonstrations of optical clocks have relied on the exceptional performance gained through stabilization using bulk cavities, which unfortunately necessitates the use of vacuum and also renders the laser susceptible to vibration-induced noise. Here, using a stimulated Brillouin scattering laser subsystem that has a reduced cavity volume and operates without vacuum, we demonstrate a promising component of a portable optical atomic clock architecture. We interrogate a 88Sr+ ion with our stimulated Brillouin scattering laser and achieve a clock exhibiting short-term stability of 3.9 × 10−14 over one second—an improvement of an order of magnitude over state-of-the-art microwave clocks. This performance increase within a potentially portable system presents a compelling avenue for substantially improving existing technology, such as the global positioning system, and also for enabling the exploration of topics such as geodetic measurements of the Earth, searches for dark matter and investigations into possible long-term variations of fundamental physics constants.
Summary
Microwave atomic clocks have traditionally served as the 'gold standard' for precision measurements of time and frequency. However, over the past decade, optical atomic clocks have surpassed the precision of their microwave counterparts by two orders of magnitude or more. Extant optical clocks occupy volumes of more than one cubic...
READ MORE
Adaptive stress testing: finding likely failure events with reinforcement learning
December 1, 2020
Journal Article
Published in:
J. Artif. Intell. Res., Vol. 69, 2020, pp. 1165-1201.
Summary
Finding the most likely path to a set of failure states is important to the analysis of safety critical systems that operate over a sequence of time steps, such as aircraft collision avoidance systems and autonomous cars. In many applications such as autonomous driving, failures cannot be completely eliminated due to the complex stochastic environment in which the system operates. As a result, safety validation is not only concerned about whether a failure can occur, but also discovering which failures are most likely to occur. This article presents adaptive stress testing (AST), a framework for finding the most likely path to a failure event in simulation. We consider a general black box setting for partially observable and continuous-valued systems operating in an environment with stochastic disturbances. We formulate the problem as a Markov decision process and use reinforcement learning to optimize it. The approach is simulation-based and does not require internal knowledge of the system, making it suitable for black-box testing of large systems. We present different formulations depending on whether the state is fully observable or partially observable. In the latter case, we present a modified Monte Carlo tree search algorithm that only requires access to the pseudorandom number generator of the simulator to overcome partial observability. We also present an extension of the framework, called differential adaptive stress testing (DAST), that can find failures that occur in one system but not in another. This type of differential analysis is useful in applications such as regression testing, where we are concerned with finding areas of relative weakness compared to a baseline. We demonstrate the effectiveness of the approach on an aircraft collision avoidance application, where a prototype aircraft collision avoidance system is stress tested to find the most likely scenarios of near mid-air collision.
Summary
Finding the most likely path to a set of failure states is important to the analysis of safety critical systems that operate over a sequence of time steps, such as aircraft collision avoidance systems and autonomous cars. In many applications such as autonomous driving, failures cannot be completely eliminated due...
READ MORE
Ultrasound diagnosis of COVID-19: robustness and explainability
November 30, 2020
Journal Article
Published in:
arXiv:2012.01145v1 [eess.IV]
Topic:
R&D area:
Summary
Diagnosis of COVID-19 at point of care is vital to the containment of the global pandemic. Point of care ultrasound (POCUS) provides rapid imagery of lungs to detect COVID-19 in patients in a repeatable and cost effective way. Previous work has used public datasets of POCUS videos to train an AI model for diagnosis that obtains high sensitivity. Due to the high stakes application we propose the use of robust and explainable techniques. We demonstrate experimentally that robust models have more stable predictions and offer improved interpretability. A framework of contrastive explanations based on adversarial perturbations is used to explain model predictions that aligns with human visual perception.
Summary
Diagnosis of COVID-19 at point of care is vital to the containment of the global pandemic. Point of care ultrasound (POCUS) provides rapid imagery of lungs to detect COVID-19 in patients in a repeatable and cost effective way. Previous work has used public datasets of POCUS videos to train an...
READ MORE