Search

We present a noisy channel generative model of two sequences, for example text and speech, which enables uncovering the association between the two modalities when limited paired data is available. To address the intractability of the exact model under a realistic data setup, we propose a variational inference approximation. To train this variational model with categorical data, we propose a KL encoder loss approach which has connections to the wake-sleep algorithm. Identifying the joint or conditional distributions by only observing unpaired samples from the marginals is only possible under certain conditions in the data distribution and we discuss under what type of conditional independence assumptions that might be achieved, which guides the architecture designs. Experimental results show that even tiny amount of paired data (5 minutes) is sufficient to learn to relate the two modalities (graphemes and phonemes here) when a massive amount of unpaired data is available, paving the path to adopting this principled approach for all seq2seq models in low data resource regimes.

Background: The SARS-CoV-2 virus has infected millions of people, overwhelming critical care resources in some regions. Many plans for rationing critical care resources during crises are based on the Sequential Organ Failure Assessment (SOFA) score. The COVID-19 pandemic created an emergent need to develop and validate a novel electronic health record (EHR)-computable tool to predict mortality. Research Questions: To rapidly develop, validate, and implement a novel real-time mortality score for the COVID-19 pandemic that improves upon SOFA. Study Design and Methods: We conducted a prospective cohort study of a regional health system with 12 hospitals in Colorado between March 2020 and July 2020. All patients >14 years old hospitalized during the study period without a do not resuscitate order were included. Patients were stratified by the diagnosis of COVID-19. From this cohort, we developed and validated a model using stacked generalization to predict mortality using data widely available in the EHR by combining five previously validated scores and additional novel variables reported to be associated with COVID-19-specific mortality. We compared the area under the receiver operator curve (AUROC) for the new model to the SOFA score and the Charlson Comorbidity Index. Results: We prospectively analyzed 27,296 encounters, of which 1,358 (5.0%) were positive for SARS-CoV-2, 4,494 (16.5%) included intensive care unit (ICU)-level care, 1,480 (5.4%) included invasive mechanical ventilation, and 717 (2.6%) ended in death. The Charlson Comorbidity Index and SOFA scores predicted overall mortality with an AUROC of 0.72 and 0.90, respectively. Our novel score predicted overall mortality with AUROC 0.94. In the subset of patients with COVID-19, we predicted mortality with AUROC 0.90, whereas SOFA had AUROC of 0.85. Interpretation: We developed and validated an accurate, in-hospital mortality prediction score in a live EHR for automatic and continuous calculation using a novel

BackgroundSince the onset of the SARS-CoV-2 pandemic, most clinical testing has focused on RT-PCR1. Host epigenome manipulation post coronavirus infection2-4 suggests that DNA methylation signatures may differentiate patients with SARS-CoV-2 infection from uninfected individuals, and help predict COVID-19 disease severity, even at initial presentation.MethodsWe customized Illumina's Infinium MethylationEPIC array to enhance immune response detection and profiled peripheral blood samples from 164 COVID-19 patients with longitudinal measurements of disease severity and 296 patient controls.ResultsEpigenome-wide association analysis revealed 13,033 genome-wide significant methylation sites for case-vs-control status. Genes and pathways involved in interferon signaling and viral response were significantly enriched among differentially methylated sites. We observe highly significant associations at genes previously reported in genetic association studies (e.g. IRF7, OAS1). Using machine learning techniques, models built using sparse regression yielded highly predictive findings: cross-validated best fit AUC was 93.6% for case-vs-control status, and 79.1%, 80.8%, and 84.4% for hospitalization, ICU admission, and progression to death, respectively.ConclusionsIn summary, the strong COVID-19-specific epigenetic signature in peripheral blood driven by key immune-related pathways related to infection status, disease severity, and clinical deterioration provides insights useful for diagnosis and prognosis of patients with viral infections.

This work explores the task of synthesizing speech in nonexistent human-sounding voices. We call this task "speaker generation", and present TacoSpawn, a system that performs competitively at this task. TacoSpawn is a recurrent attention-based text-to-speech model that learns a distribution over a speaker embedding space, which enables sampling of novel and diverse speakers. Our method is easy to implement, and does not require transfer learning from speaker ID systems. We present objective and subjective metrics for evaluating performance on this task, and demonstrate that our proposed objective metrics correlate with human perception of speaker similarity. Audio samples are available on our demo page., Comment: 12 pages, 3 figures, 4 tables, appendix with 2 tables

The recent development of data-driven AI promises to automate medical diagnosis; however, most AI functions as 'black boxes' to physicians with limited computational knowledge. Using medical imaging as a point of departure, we conducted three iterations of design activities to formulate CheXplain---a system that enables physicians to explore and understand AI-enabled chest X-ray analysis: (1) a paired survey between referring physicians and radiologists reveals whether, when, and what kinds of explanations are needed; (2) a low-fidelity prototype co-designed with three physicians formulates eight key features; and (3) a high-fidelity prototype evaluated by another six physicians provides detailed summative insights on how each feature enables the exploration and understanding of AI. We summarize by discussing recommendations for future work to design and implement explainable medical AI systems that encompass four recurring themes: motivation, constraint, explanation, and justification., Comment: 10 pages, 5 figures

Non-saturating generative adversarial network (GAN) training is widely used and has continued to obtain groundbreaking results. However so far this approach has lacked strong theoretical justification, in contrast to alternatives such as f-GANs and Wasserstein GANs which are motivated in terms of approximate divergence minimization. In this paper we show that non-saturating GAN training does in fact approximately minimize a particular f-divergence. We develop general theoretical tools to compare and classify f-divergences and use these to show that the new f-divergence is qualitatively similar to reverse KL. These results help to explain the high sample quality but poor diversity often observed empirically when using this scheme.

Despite the ability to produce human-level speech for in-domain text, attention-based end-to-end text-to-speech (TTS) systems suffer from text alignment failures that increase in frequency for out-of-domain text. We show that these failures can be addressed using simple location-relative attention mechanisms that do away with content-based query/key comparisons. We compare two families of attention mechanisms: location-relative GMM-based mechanisms and additive energy-based mechanisms. We suggest simple modifications to GMM-based attention that allow it to align quickly and consistently during training, and introduce a new location-relative attention mechanism to the additive energy-based family, called Dynamic Convolution Attention (DCA). We compare the various mechanisms in terms of alignment speed and consistency during training, naturalness, and ability to generalize to long utterances, and conclude that GMM attention and DCA can generalize to very long utterances, while preserving naturalness for shorter, in-domain utterances., Comment: Accepted to ICASSP 2020

We present a novel generative model that combines state-of-the-art neural text-to-speech (TTS) with semi-supervised probabilistic latent variable models. By providing partial supervision to some of the latent variables, we are able to force them to take on consistent and interpretable purposes, which previously hasn't been possible with purely unsupervised TTS models. We demonstrate that our model is able to reliably discover and control important but rarely labelled attributes of speech, such as affect and speaking rate, with as little as 1% (30 minutes) supervision. Even at such low supervision levels we do not observe a degradation of synthesis quality compared to a state-of-the-art baseline. Audio samples are available on the web.

Recent work has explored sequence-to-sequence latent variable models for expressive speech synthesis (supporting control and transfer of prosody and style), but has not presented a coherent framework for understanding the trade-offs between the competing methods. In this paper, we propose embedding capacity (the amount of information the embedding contains about the data) as a unified method of analyzing the behavior of latent variable models of speech, comparing existing heuristic (non-variational) methods to variational methods that are able to explicitly constrain capacity using an upper bound on representational mutual information. In our proposed model (Capacitron), we show that by adding conditional dependencies to the variational posterior such that it matches the form of the true posterior, the same model can be used for high-precision prosody transfer, text-agnostic style transfer, and generation of natural-sounding prior samples. For multi-speaker models, Capacitron is able to preserve target speaker identity during inter-speaker prosody transfer and when drawing samples from the latent prior. Lastly, we introduce a method for decomposing embedding capacity hierarchically across two sets of latents, allowing a portion of the latent variability to be specified and the remaining variability sampled from a learned prior. Audio examples are available on the web., Comment: Submitted to ICLR 2020

Astrophysics is transforming from a data-starved to a data-swamped discipline, fundamentally changing the nature of scientific inquiry and discovery. New technologies are enabling the detection, transmission, and storage of data of hitherto unimaginable quantity and quality across the electromagnetic, gravity and particle spectra. The observational data obtained during this decade alone will supersede everything accumulated over the preceding four thousand years of astronomy. Currently there are 4 large-scale photometric and spectroscopic surveys underway, each generating and/or utilizing hundreds of terabytes of data per year. Some will focus on the static universe while others will greatly expand our knowledge of transient phenomena. Maximizing the science from these programs requires integrating the processing pipeline with high-performance computing resources. These are coupled to large astrophysics databases while making use of machine learning algorithms with near real-time turnaround. Here we present an overview of one of these programs, the Palomar Transient Factory (PTF). We will cover the processing and discovery pipeline we developed at LBNL and NERSC for it and several of the great discoveries made during the 4 years of observations with PTF.

We introduce the blind index coding (BIC) problem, in which a single sender communicates distinct messages to multiple users over a shared channel. Each user has partial knowledge of each message as side information. However, unlike classic index coding, in BIC, the sender is uncertain of what side information is available to each user. In particular, the sender only knows the amount of bits in each user's side information but not its content. This problem can arise naturally in caching and wireless networks. In order to blindly exploit side information in the BIC problem, we develop a hybrid coding scheme that XORs uncoded bits of a subset of messages with random combinations of bits from other messages. This scheme allows us to strike the right balance between maximizing the transmission rate to each user and minimizing the interference leakage to others. We also develop a general outer bound, which relies on a strong data processing inequality to effectively capture the senders uncertainty about the users' side information. Additionally, we consider the case where communication takes place over a shared wireless medium, modeled by an erasure broadcast channel, and show that surprisingly, combining repetition coding with hybrid coding improves the achievable rate region and outperforms alternative strategies of coping with channel erasure and while blindly exploiting side information., Comment: Parts of this paper were presented at ISIT 2015 and ICC 2015

In this paper, we study the communication problem from rovers on Mars' surface to Mars-orbiting satellites. We first justify that, to a good extent, the rover-to-orbiter communication problem can be modelled as communication over a $2 \times 2$ X-channel with the network topology varying over time. For such a fading X-channel where transmitters are only aware of the time-varying topology but not the time-varying channel state (i.e., no CSIT), we propose coding strategies that code across topologies, and develop upper bounds on the sum degrees-of-freedom (DoF) that is shown to be tight under certain pattern of the topology variation. Furthermore we demonstrate that the proposed scheme approximately achieves the ergodic sum-capacity of the network. Using the proposed coding scheme, we numerically evaluate the ergodic rate gain over a time-division-multiple-access (TDMA) scheme for Rayleigh and Rice fading channels. We also numerically demonstrate that with practical orbital parameters, a 9.6% DoF gain, as well as more than 11.6% throughput gain can be achieved for a rover-to-orbiter communication network., Comment: 13 pages, 6 figures. Accepted by IEEE Transactions on Communications

We study the degrees of freedom (DoF) of the multiple-input multiple-output X-channel (MIMO XC) with delayed channel state information at the transmitters (delayed CSIT), assuming linear coding strategies at the transmitters. We present two results: 1) the linear sum DoF for MIMO XC with general antenna configurations, and 2) the linear DoF region for MIMO XC with symmetric antennas. The converse for each result is based on developing a novel rank-ratio inequality that characterizes the maximum ratio between the dimensions of received linear subspaces at the two multiple-antenna receivers. The achievability of the linear sum DoF is based on a three-phase strategy, in which during the first two phases only the transmitter with fewer antennas exploits delayed CSIT in order to minimize the dimension of its signal at the unintended receiver. During Phase 3, both transmitters use delayed CSIT to send linear combinations of past transmissions such that each receiver receives a superposition of desired message data and known interference, thus simultaneously serving both receivers. We also derive other linear DoF outer bounds for the MIMO XC that, in addition to the outer bounds from the sum DoF converse and the proposed transmission strategy, allow us to characterize the linear DoF region for symmetric antenna configurations., Comment: to be presented in part at ISIT 2014

Dirty paper codes are a powerful tool for combating known interference. However, there is a significant difference between knowing the transmitted interference sequence and knowing the received interference sequence, especially when the channel modifying the interference is uncertain. We present an upper bound on the capacity of a compound vector dirty paper channel where although an additive Gaussian sequence is known to the transmitter, the channel matrix between the interferer and receiver is uncertain but known to lie within a bounded set. Our bound is tighter than previous bounds in the low-SIR regime for the scalar version of the compound dirty paper channel and employs a construction that focuses on the relationship between the dimension of the message-bearing signal and the dimension of the additive state sequence. Additionally, a bound on the high-SNR behavior of the system is established., Comment: to be presented at ISIT 2013

We study the capacity regions of two-user interference channels where transmitters base their transmission schemes on local views of the channel state. Under the local view model, each transmitter knows only a subset of the four channel gains, which may be mismatched from the other transmitter. We consider a set of seven local views, and find that for five out of the seven local views, TDM is sufficient to achieve the qualified notion of capacity region for the linear deterministic interference channel which approximates the Gaussian interference channel. For these five local views, the qualified capacity result implies that no policy can achieve a rate point outside the TDM region without inducing a corner case of sub-TDM performance in another channel state. The common trait shared by the two remaining local views - those with the potential to outperform TDM - is transmitter knowledge of the outgoing interference link accompanied by some common knowledge of state, emphasizing their importance in creating opportunities to coordinate usage of more advanced schemes. Our conclusions are extended to bounded gap characterizations of the capacity region for the Gaussian interference channel., Comment: revised 22 Jun, including updated title

We present a set of five axioms for fairness measures in resource allocation. A family of fairness measures satisfying the axioms is constructed. Well-known notions such as alpha-fairness, Jain's index, and entropy are shown to be special cases. Properties of fairness measures satisfying the axioms are proven, including Schur-concavity. Among the engineering implications is a generalized Jain's index that tunes the resolution of the fairness measure, a new understanding of alpha-fair utility functions, and an interpretation of "larger alpha is more fair". We also construct an alternative set of four axioms to capture efficiency objectives and feasibility constraints.

This thesis examines the use of the Integrated Product Team (IPT) concept within the Tomahawk Cruise Missile Program at the Defense Contracting Management Agency (DCMA) Raytheon. The study presents a literature review of the IPT philosophy and concepts and an overview of the Tomahawk Cruise Missile Program IPTs. Surveys and interviews focus on the following areas: 1) PT dynamics; 2) IPT performance; 3) IPT training; and 4) the working relationship between contractor and Government personnel on lPTs. Overall, team members who have participated in or managed an IPT agree that PT has added value to the acquisition process by bringing functional disciplines from Government and industry together to exchange ideas and build a successful Tomahawk Program. Only a few team members reported that the IPT process led to problems in the decision-making process, alignment of team objectives, and contractor and Government working relationships. Although both contractor and Government team members have some misunderstandings and preconceived notions about each other, both organizations realize that an effective Government and contractor interface provided by the PT process is crucial to the success of the Tomahawk Program.

This thesis examines the use of the Integrated Product Team (IPT) concept within the Tomahawk Cruise Missile Program at the Defense Contracting Management Agency (DCMA) Raytheon. The study presents a literature review of the IPT philosophy and concepts and an overview of the Tomahawk Cruise Missile Program IPTs. Surveys and interviews focus on the following areas: 1) PT dynamics; 2) IPT performance; 3) IPT training; and 4) the working relationship between contractor and Government personnel on lPTs. Overall, team members who have participated in or managed an IPT agree that PT has added value to the acquisition process by bringing functional disciplines from Government and industry together to exchange ideas and build a successful Tomahawk Program. Only a few team members reported that the IPT process led to problems in the decision-making process, alignment of team objectives, and contractor and Government working relationships. Although both contractor and Government team members have some misunderstandings and preconceived notions about each other, both organizations realize that an effective Government and contractor interface provided by the PT process is crucial to the success of the Tomahawk Program.

This thesis examines the use of the Integrated Product Team (IPT) concept within the Tomahawk Cruise Missile Program at the Defense Contracting Management Agency (DCMA) Raytheon. The study presents a literature review of the IPT philosophy and concepts and an overview of the Tomahawk Cruise Missile Program IPTs. Surveys and interviews focus on the following areas: 1) PT dynamics; 2) IPT performance; 3) IPT training; and 4) the working relationship between contractor and Government personnel on lPTs. Overall, team members who have participated in or managed an IPT agree that PT has added value to the acquisition process by bringing functional disciplines from Government and industry together to exchange ideas and build a successful Tomahawk Program. Only a few team members reported that the IPT process led to problems in the decision-making process, alignment of team objectives, and contractor and Government working relationships. Although both contractor and Government team members have some misunderstandings and preconceived notions about each other, both organizations realize that an effective Government and contractor interface provided by the PT process is crucial to the success of the Tomahawk Program., http://archive.org/details/theuseofintegrat109459192, U.S. Navy (U.S.N.) author., Approved for public release; distribution is unlimited.