Showing papers by "Neal Patwari published in 2022"

Models for understanding and quantifying feedback in societal systems

Abstract: When it comes to long-term fairness in decision-making settings, many studies have focused on closed systems with a specific appointed decision-maker and certain engagement rules in place. However, if the objective is to achieve equity in a broader societal system, studying the system in isolation is insufficient. In a societal system, neither a singular decision maker nor defined agent behavior rules exist. Additionally, analysis of societal systems can be complicated by the presence of feedback, in which historical and current inequities influence future inequity. In this paper, we present a model to quantify feedback in social systems so that the long-term effects of a policy or decision process may be investigated, even when the feedback mechanisms are not individually characterized. We explore the dynamics of real social systems and find that many examples of feedback are qualitatively similar in their temporal characteristics. Using a key idea in linear systems theory, namely proportional-integral-derivative (PID) feedback, we propose a model to quantify three types of feedback. We illustrate how different components of the PID capture analogous aspects of societal dynamics such as the persistence of current inequity, the cumulative effects of long-term inequity, and the response to the speed at which society is changing. Our model does not attempt to describe underlying systems or capture individual actions. It is a system-based approach to study inequity in feedback loops, and as a result unlocks a direction to study social systems that would otherwise be almost impossible to model and can only be observed. Our framework helps elucidate the ability of fair policies to produce and sustain equity in the long-term.

Deep Learning-based Localization in Limited Data Regimes

Abstract: As demand for radio spectrum increases with the widespread use of wireless devices, effective spectrum allocation requires more flexibility in terms of time, space, and frequency. In order to protect users in next-generation wireless networks from interference, spectrum managers must have the ability to efficiently and accurately locate transmitters. We present TL;DL, a practical deep-learning based technique for multiple transmitter localization on crowdsourced data where all sensors and transmitters may be mobile and transmit with unknown power. We map sensor readings to an image representing the sensor location, then use a convolutional neural network to learn to generate a target image of transmitter locations. We also introduce a novel data-augmentation technique to drastically improve generalization and enable accurate localization on limited data. In our evaluation, TL;DL outperforms previous approaches on small real-world datasets with low sensor density, in terms of both accuracy and detection.

Pseudonymetry: Precise, Private Closed Loop Control for Spectrum Reuse with Passive Receivers

Abstract: Radio astronomy and other passive radio spectrum users have significant challenges avoiding interference from wireless communication systems. Even distant transmitters sometimes interfere with passive users. We propose Pseudonymetry, a system that provides (primary) passive users a means to turn off the transmissions of the particular (secondary) wireless transmitter that interferes with it. By controlling the specific transmitter rather than an entire geographical region, Pseudonymetry could increase the spectrum available for wireless systems while ensuring rapid clearing of interferers as necessary for passive use. Pseudonymetry adds a low rate watermark to the secondary (intended) transmitted signal to carry a random, anonymous pseudonym. We show the ability of a passive receiver to decode the watermark, even from a signal received with very low SNR. The passive receiver posts to a centralized database to provide feedback to the secondary transmitters so that they know to vacate the band. We provide analysis that captures the trade-offs in the design of Pseudonymetry, and show initial evidence that a simple amplitude modulation watermarking scheme could enable reliable detection at a distant passive receiver, while resulting in minimal degradation to the error performance of the intended secondary receiver.

Digital Spectrum Twinning for Next-Generation Spectrum Management and Metering

Abstract: This paper further develops the concept of a digital spectrum twin (DST), which can be used to enhance dynamic radio access and spectrum management. Specifically, we define the three types of parallel intelligence at work in a DST. With this framework, we demonstrate with economic principles and an illustrative case study the importance of spectrum metering on a DST-enabled radioscape.

Short: Racial Disparities in Pulse Oximetry Cannot Be Fixed With Race-Based Correction

Abstract: Pulse oximeters play a critical role in health monitoring. Pulse ox measurements have statistical bias that is a function of race, which results in higher rates of occult hypoxemia, i.e., missed detection of dangerously low oxygenation, in patients of color. This paper further characterizes the statistical distribution of pulse ox measurements, showing they also have a higher variance for patients racialized as Black, compared to those racialized as white. By analyzing the performance of hypoxemia detection as a detector, we show that no single race-based correction factor will provide equal performance by race. As a result, for racially equitable pulse oximetry, the pulse oximeter itself must be fixed, not just the hypoxemia thresholds.

Two Measure is Two Know: Calibration-free Full Duplex Monitoring for Software Radio Platforms

Abstract: Future virtualized radio access network (vRAN) infrastructure providers (and today’s experimental wireless testbed providers) may be simultaneously uncertain what signals are being transmitted by their base stations and legally responsible for their violations. These providers must monitor the spectrum of transmissions and external signals without access to the radio itself. In this paper, we propose FDMonitor, a full-duplex monitoring system attached between a transmitter and its antenna to achieve this goal. Measuring the signal at this point on the RF path is necessary but insufficient since the antenna is a bidirectional device. FDMonitor thus uses a bidirectional coupler, a two-channel receiver, and a new source separation algorithm to simultaneously estimate the transmitted signal and the signal incident on the antenna. Rather than requiring an offline calibration, we also adaptively estimate the linear model for the system on the fly. FDMonitor has been running on a real-world open wireless testbed, monitoring 19 SDR platforms controlled (with bare metal access) by outside experimenters over a seven month period, sending alerts whenever a violation is observed. Our experimental results show that FDMonitor accurately separates signals across a range of signal param-eters. Over more than 7 months of observation, it achieves a positive predictive value of 97%, with a total of 20 false alerts

Global Performance Disparities Between English-Language Accents in Automatic Speech Recognition

Abstract: Past research has identified discriminatory automatic speech recognition (ASR) performance as a function of the racial group and nationality of the speaker. In this paper, we expand the discussion beyond bias as a function of the individual national origin of the speaker to look for bias as a function of the geopolitical orientation of their nation of origin. We audit some of the most popular English language ASR services using a large and global data set of speech from The Speech Accent Archive, which includes over 2,700 speakers of English born in 171 different countries. We show that, even when controlling for multiple linguistic covariates, ASR service performance has a statistically significant relationship to the political alignment of the speaker's birth country with respect to the United States' geopolitical power. This holds for all ASR services tested. We discuss this bias in the context of the historical use of language to maintain global and political power.