More Like this

1. Coverage Path Planning for Mapping of Underwater Structures

   https://doi.org/10.1109/IEEECONF38699.2020.9389389  

   Karapetyan, Nare; Johnson, James V.; Rekleitis, Ioannis (October 2020, Global Oceans 2020: Singapore – U.S. Gulf Coast)

   null (Ed.)

   This paper addresses the problem of the coverage path planning in a 3D environment for surveying underwater structures. We propose to use the navigation strategy that a human diver will execute when circumnavigating around a region of interest, in particular when collecting data from a shipwreck. In contrast to the previous methods in the literature, we are aiming to perform coverage in completely unknown environment with some initial prior information. Our proposed method uses convolutional neural networks to learn the control commands based on the visual input. Preliminary results and a detailed overview of the proposed method are discussed. 

   more » « less
2. Direction Finding and Likelihood Ratio Detection for Oceanographic HF Radars

   https://doi.org/10.1175/JTECH-D-21-0110.1  

   Emery, Brian; Kirincich, Anthony; Washburn, Libe (February 2022, Journal of Atmospheric and Oceanic Technology)

   Abstract Previous work with simulations of oceanographic high-frequency (HF) radars has identified possible improvements when using maximum likelihood estimation (MLE) for direction of arrival; however, methods for determining the number of emitters (here defined as spatially distinct patches of the ocean surface) have not realized these improvements. Here we describe and evaluate the use of the likelihood ratio (LR) for emitter detection, demonstrating its application to oceanographic HF radar data. The combined detection–estimation methods MLE-LR are compared with multiple signal classification method (MUSIC) and MUSIC parameters for SeaSonde HF radars, along with a method developed for 8-channel systems known as MUSIC-Highest. Results show that the use of MLE-LR produces similar accuracy, in terms of the RMS difference and correlation coefficients squared, as previous methods. We demonstrate that improved accuracy can be obtained for both methods, at the cost of fewer velocity observations and decreased spatial coverage. For SeaSondes, accuracy improvements are obtained with less commonly used parameter sets. The MLE-LR is shown to be able to resolve simultaneous closely spaced emitters, which has the potential to improve observations obtained by HF radars operating in complex current environments. Significance Statement We identify and test a method based on the likelihood ratio (LR) for determining the number of signal sources in observations subject to direction finding with maximum likelihood estimation (MLE). Direction-finding methods are used in broad-ranging applications that include radar, sonar, and wireless communication. Previous work suggests accuracy improvements when using MLE, but suitable methods for determining the number of simultaneous signal sources are not well known. Our work shows that the LR, when combined with MLE, performs at least as well as alternative methods when applied to oceanographic high-frequency (HF) radars. In some situations, MLE and LR obtain superior resolution, where resolution is defined as the ability to distinguish closely spaced signal sources. 

   more » « less
3. Learning Bellman Complete Representations for Offline Policy Evaluation

   Jonathan Chang, Kaiwen Wang (January 2022, Proceedings of the 39th International Conference on Machine Learning)

   We study representation learning for Offline Reinforcement Learning (RL), focusing on the important task of Offline Policy Evaluation (OPE). Recent work shows that, in contrast to supervised learning, realizability of the Q-function is not enough for learning it. Two sufficient conditions for sample-efficient OPE are Bellman completeness and coverage. Prior work often assumes that representations satisfying these conditions are given, with results being mostly theoretical in nature. In this work, we propose BCRL, which directly learns from data an approximately linear Bellman complete representation with good coverage. With this learned representation, we perform OPE using Least Square Policy Evaluation (LSPE) with linear functions in our learned representation. We present an end-to-end theoretical analysis, showing that our two-stage algorithm enjoys polynomial sample complexity provided some representation in the rich class considered is linear Bellman complete. Empirically, we extensively evaluate our algorithm on challenging, image-based continuous control tasks from the Deepmind Control Suite. We show our representation enables better OPE compared to previous representation learning methods developed for off-policy RL (e.g., CURL, SPR). BCRL achieve competitive OPE error with the state-of-the-art method Fitted Q-Evaluation (FQE), and beats FQE when evaluating beyond the initial state distribution. Our ablations show that both linear Bellman complete and coverage components of our method are crucial. 

   more » « less
4. Phertilizer: Growing a clonal tree from ultra-low coverage single-cell DNA sequencing of tumors

   https://doi.org/10.1371/journal.pcbi.1011544  

   Weber, Leah L.; Zhang, Chuanyi; Ochoa, Idoia; El-Kebir, Mohammed (October 2023, PLOS Computational Biology)

   Przytycka, Teresa M. (Ed.)

   Emerging ultra-low coverage single-cell DNA sequencing (scDNA-seq) technologies have enabled high resolution evolutionary studies of copy number aberrations (CNAs) within tumors. While these sequencing technologies are well suited for identifying CNAs due to the uniformity of sequencing coverage, the sparsity of coverage poses challenges for the study of single-nucleotide variants (SNVs). In order to maximize the utility of increasingly available ultra-low coverage scDNA-seq data and obtain a comprehensive understanding of tumor evolution, it is important to also analyze the evolution of SNVs from the same set of tumor cells. We presentPhertilizer, a method to infer a clonal tree from ultra-low coverage scDNA-seq data of a tumor. Based on a probabilistic model, our method recursively partitions the data by identifying key evolutionary events in the history of the tumor. We demonstrate the performance ofPhertilizeron simulated data as well as on two real datasets, finding thatPhertilizereffectively utilizes the copy-number signal inherent in the data to more accurately uncover clonal structure and genotypes compared to previous methods. 

   more » « less
5. Automated Coverage-Based Testing of XACML Policies

   https://doi.org/10.1145/3205977.3205979  

   Xu, Dianxiang; Shrestha, Roshan; Shen, Ning (June 2018, Proc. of the 23rd ACM Symposium on Access Control Models and Technologies)

   While the standard language XACML is very expressive for specifying fine-grained access control policies, defects can get into XACML policies for various reasons, such as misunderstanding of access control requirements, omissions, and coding errors. These defects may result in unauthorized accesses, escalation of privileges, and denial of service. Therefore, quality assurance of XACML policies for real-world information systems has become an important issue. To address this issue, this paper presents a family of coverage criteria for XACML policies, such as rule coverage, rule pair coverage, decision coverage, and Modified Condition/Decision Coverage (MC/DC). To demonstrate the assurance levels of these coverage criteria, we have developed methods for automatically generating tests, i.e., access requests, to satisfy the coverage criteria using a constraint solver. We have evaluated these methods through mutation analysis of various policies with different levels of complexity. The experiment results have shown that the rule coverage is far from adequate for revealing the majority of defects in XACML policies, and that both MC/DC and decision coverage tests have outperformed the existing methods for testing XACML policies. In particular, MC/DC tests achieve a very high level of quality assurance of XACML policies. 

   more » « less