arXiv subjects

Yuhao Zhang

Publications and source records attributed to Yuhao Zhang.

At least 145 records · Page 8Linked to original sources

arXiv2023

Fair $k$-Center: a Coreset Approach in Low Dimensions

Center-based clustering techniques are fundamental in some areas of machine learning such as data summarization. Generic $k$-center algorithms can produce biased cluster representatives so there has been a recent interest in fair $k$-center clustering. Our main theoretical contributions are two new $(3+ε)$-approximation algorithms for solving the fair $k$-center problem in (1) the dynamic incremental, i.e., one-pass streaming, model and (2) the MapReduce model. Our dynamic incremental algorithm is the first such algorithm for this problem (previous streaming algorithms required two passes) and our MapReduce one improves upon the previous approximation factor of $(17+ε).$ Both algorithms work by maintaining a small coreset to represent the full point set and their analysis requires that the underlying metric has finite-doubling dimension. We also provide related heuristics for higher dimensional data and experimental results that compare the performance of our algorithms to existing ones.

cs.DC↗

arXiv2022

Improving Cross-task Generalization of Unified Table-to-text Models with Compositional Task Configurations

There has been great progress in unifying various table-to-text tasks using a single encoder-decoder model trained via multi-task learning (Xie et al., 2022). However, existing methods typically encode task information with a simple dataset name as a prefix to the encoder. This not only limits the effectiveness of multi-task learning, but also hinders the model's ability to generalize to new domains or tasks that were not seen during training, which is crucial for real-world applications. In this paper, we propose compositional task configurations, a set of prompts prepended to the encoder to improve cross-task generalization of unified models. We design the task configurations to explicitly specify the task type, as well as its input and output types. We show that this not only allows the model to better learn shared knowledge across different tasks at training, but also allows us to control the model by composing new configurations that apply novel input-output combinations in a zero-shot manner. We demonstrate via experiments over ten table-to-text tasks that our method outperforms the UnifiedSKG baseline by noticeable margins in both in-domain and zero-shot settings, with average improvements of +0.5 and +12.6 from using a T5-large backbone, respectively.

cs.CL↗

arXiv2022

Improving End-to-end Speech Translation by Leveraging Auxiliary Speech and Text Data

We present a method for introducing a text encoder into pre-trained end-to-end speech translation systems. It enhances the ability of adapting one modality (i.e., source-language speech) to another (i.e., source-language text). Thus, the speech translation model can learn from both unlabeled and labeled data, especially when the source-language text data is abundant. Beyond this, we present a denoising method to build a robust text encoder that can deal with both normal and noisy text data. Our system sets new state-of-the-arts on the MuST-C En-De, En-Fr, and LibriSpeech En-Fr tasks.

eess.AS↗

arXiv2022

Thermal Modulation of Gigahertz Surface Acoustic Waves on Lithium Niobate

Surface acoustic wave (SAW) devices have wide range of applications in microwave signal processing. Microwave SAW components benefit from higher quality factors and much smaller crosstalk when compared to their electromagnetic counterparts. Efficient routing and modulation of SAWs are essential for building large-scale and versatile acoustic-wave circuits. Here, we demonstrate integrated thermo-acoustic modulators using two SAW platforms: bulk lithium niobate and thin-film lithium niobate on sapphire. In both approaches, the gigahertz-frequency SAWs are routed by integrated acoustic waveguides while on-chip microheaters are used to locally change the temperature and thus control the phase of SAW. Using this approach, we achieved phase changes of over 720 degrees with the responsibility of 2.6 deg/mW for bulk lithium niobate and 0.52 deg/mW for lithium niobate on sapphire. Furthermore, we demonstrated amplitude modulation of SAWs using acoustic Mach Zehnder interferometers. Our thermo-acoustic modulators can enable reconfigurable acoustic signal processing for next generation wireless communications and microwave systems.

physics.app-ph↗

arXiv2022

BagFlip: A Certified Defense against Data Poisoning

Machine learning models are vulnerable to data-poisoning attacks, in which an attacker maliciously modifies the training set to change the prediction of a learned model. In a trigger-less attack, the attacker can modify the training set but not the test inputs, while in a backdoor attack the attacker can also modify test inputs. Existing model-agnostic defense approaches either cannot handle backdoor attacks or do not provide effective certificates (i.e., a proof of a defense). We present BagFlip, a model-agnostic certified approach that can effectively defend against both trigger-less and backdoor attacks. We evaluate BagFlip on image classification and malware detection datasets. BagFlip is equal to or more effective than the state-of-the-art approaches for trigger-less attacks and more effective than the state-of-the-art approaches for backdoor attacks.

cs.LG↗

arXiv2022

Safety Verification of Neural Feedback Systems Based on Constrained Zonotopes

Artificial neural networks have recently been utilized in many feedback control systems and introduced new challenges regarding the safety of such systems. This paper considers the safe verification problem for a dynamical system with a given feedforward neural network as the feedback controller by using a constrained zonotope-based approach. A novel set-based method is proposed to compute both exact and over-approximated reachable sets for neural feedback systems with linear models, and linear program-based sufficient conditions are presented to verify whether the trajectories of such a system can avoid unsafe regions represented as constrained zonotopes. The results are also extended to neural feedback systems with nonlinear models. The computational efficiency and accuracy of the proposed method are demonstrated by two numerical examples where a comparison with state-of-the-art methods is also provided.

math.OC↗

arXiv2022

Reachability Analysis and Safety Verification of Neural Feedback Systems via Hybrid Zonotopes

Hybrid zonotopes generalize constrained zonotopes by introducing additional binary variables and possess some unique properties that make them convenient to represent nonconvex sets. This paper presents novel hybrid zonotope-based methods for the reachability analysis and safety verification of neural feedback systems. Algorithms are proposed to compute the input-output relationship of each layer of a feedforward neural network, as well as the exact reachable sets of neural feedback systems. In addition, a sufficient and necessary condition is formulated as a mixed-integer linear program to certify whether the trajectories of a neural feedback system can avoid unsafe regions. The proposed approach is shown to yield a formulation that provides the tightest convex relaxation for the reachable sets of the neural feedback system. Complexity reduction techniques for the reachable sets are developed to balance the computation efficiency and approximation accuracy. Two numerical examples demonstrate the superior performance of the proposed approach compared to other existing methods.

math.OC↗

arXiv2022

Contrastive Learning of Medical Visual Representations from Paired Images and Text

Learning visual representations of medical images (e.g., X-rays) is core to medical image understanding but its progress has been held back by the scarcity of human annotations. Existing work commonly relies on fine-tuning weights transferred from ImageNet pretraining, which is suboptimal due to drastically different image characteristics, or rule-based label extraction from the textual report data paired with medical images, which is inaccurate and hard to generalize. Meanwhile, several recent studies show exciting results from unsupervised contrastive learning from natural images, but we find these methods help little on medical images because of their high inter-class similarity. We propose ConVIRT, an alternative unsupervised strategy to learn medical visual representations by exploiting naturally occurring paired descriptive text. Our new method of pretraining medical image encoders with the paired text data via a bidirectional contrastive objective between the two modalities is domain-agnostic, and requires no additional expert input. We test ConVIRT by transferring our pretrained weights to 4 medical image classification tasks and 2 zero-shot retrieval tasks, and show that it leads to image representations that considerably outperform strong baselines in most settings. Notably, in all 4 classification tasks, our method requires only 10\% as much labeled training data as an ImageNet initialized counterpart to achieve better or comparable performance, demonstrating superior data efficiency.

cs.CV↗

arXiv2022

Planning 5G Networks for Rural Fixed Wireless Access

We study the planning of a rural 5G multi-user massive MIMO fixed wireless access system to offer fixed broadband service to homes. Specifically, we aim to determine the user limit, i.e., the maximum number of homes that can simultaneously receive target minimum bit rates (MBRs) on the downlink (DL) and on the uplink (UL) given a set of network resources and a cell radius. To compute that limit, we must understand how resources should be shared between the DL and UL and how user and stream selection, precoding and combining, and power distribution should be performed. We use block diagonalization and propose a static grouping strategy that organizes homes into fixed groups (of possibly different sizes) in the DL and UL; then we develop a simple approach to compute the user limit that we validate numerically. We study the impact of group size and show that smaller groups yield larger user limits in a 3.5 GHz band. We show how the user limit at different cell radii is impacted by the system bandwidth, the number of antennas at the base station and homes, the transmit power, and the DL and UL MBRs. Lastly, we offer insights into how the network could be operated.

cs.NI↗

arXiv2022

Overwatch: Learning Patterns in Code Edit Sequences

Integrated Development Environments (IDEs) provide tool support to automate many source code editing tasks. Traditionally, IDEs use only the spatial context, i.e., the location where the developer is editing, to generate candidate edit recommendations. However, spatial context alone is often not sufficient to confidently predict the developer's next edit, and thus IDEs generate many suggestions at a location. Therefore, IDEs generally do not actively offer suggestions and instead, the developer is usually required to click on a specific icon or menu and then select from a large list of potential suggestions. As a consequence, developers often miss the opportunity to use the tool support because they are not aware it exists or forget to use it. To better understand common patterns in developer behavior and produce better edit recommendations, we can additionally use the temporal context, i.e., the edits that a developer was recently performing. To enable edit recommendations based on temporal context, we present Overwatch, a novel technique for learning edit sequence patterns from traces of developers' edits performed in an IDE. Our experiments show that Overwatch has 78% precision and that Overwatch not only completed edits when developers missed the opportunity to use the IDE tool support but also predicted new edits that have no tool support in the IDE.

cs.PL↗

arXiv2022

Robustar: Interactive Toolbox Supporting Precise Data Annotation for Robust Vision Learning

We introduce the initial release of our software Robustar, which aims to improve the robustness of vision classification machine learning models through a data-driven perspective. Building upon the recent understanding that the lack of machine learning model's robustness is the tendency of the model's learning of spurious features, we aim to solve this problem from its root at the data perspective by removing the spurious features from the data before training. In particular, we introduce a software that helps the users to better prepare the data for training image classification models by allowing the users to annotate the spurious features at the pixel level of images. To facilitate this process, our software also leverages recent advances to help identify potential images and pixels worthy of attention and to continue the training with newly annotated data. Our software is hosted at the GitHub Repository https://github.com/HaohanWang/Robustar.

cs.CV↗

arXiv2022

Vertical GaN Diode BV Maximization through Rapid TCAD Simulation and ML-enabled Surrogate Model

In this paper, two methodologies are used to speed up the maximization of the breakdown volt-age (BV) of a vertical GaN diode that has a theoretical maximum BV of ~2100V. Firstly, we demonstrated a 5X faster accurate simulation method in Technology Computer-Aided-Design (TCAD). This allows us to find 50% more numbers of high BV (>1400V) designs at a given simulation time. Secondly, a machine learning (ML) model is developed using TCAD-generated data and used as a surrogate model for differential evolution optimization. It can inversely design an out-of-the-training-range structure with BV as high as 1887V (89% of the ideal case) compared to ~1100V designed with human domain expertise.

cs.LG↗

arXiv2022

Survivable Network Design Revisited: Group-Connectivity

In the classical survivable network design problem (SNDP), we are given an undirected graph $G=(V,E)$ with costs on edges and a connectivity requirement $k(s,t)$ for each pair of vertices. The goal is to find a minimum-cost subgraph $H\subseteq V$ such that every pair $(s,t)$ are connected by $k(s,t)$ edge or (openly) vertex disjoint paths, abbreviated as EC-SNDP and VC-SNDP, respectively. The seminal result of Jain [FOCS'98, Combinatorica'01] gives a $2$-approximation algorithm for EC-SNDP, and a decade later, an $O(k^3\log n)$-approximation algorithm for VC-SNDP, where $k$ is the largest connectivity requirement, was discovered by Chuzhoy and Khanna [FOCS'09, Theory Comput.'12]. While there is rich literature on point-to-point settings of SNDP, the viable case of connectivity between subsets is still relatively poorly understood. This paper concerns the generalization of SNDP into the subset-to-subset setting, namely Group EC-SNDP. We develop the framework, which yields the first non-trivial (true) approximation algorithm for Group EC-SNDP. Previously, only a bicriteria approximation algorithm is known for Group EC-SNDP [Chalermsook, Grandoni, and Laekhanukit, SODA'15], and a true approximation algorithm is known only for the single-source variant with connectivity requirement $k(S,T)\in\{0,1,2\}$ [Gupta, Krishnaswamy, and Ravi, SODA'10; Khandekar, Kortsarz, and Nutov, FSTTCS'09 and Theor. Comput. Sci.'12].

cs.DS↗

arXiv2022

Improved Bounds for Fractional Online Matching Problems

Online bipartite matching with one-sided arrival and its variants have been extensively studied since the seminal work of Karp, Vazirani, and Vazirani (STOC 1990). Motivated by real-life applications with dynamic market structures, e.g. ride-sharing, two generalizations of the classical one-sided arrival model are proposed to allow non-bipartite graphs and to allow all vertices to arrive online. Namely, online matching with general vertex arrival is introduced by Wang and Wong (ICALP 2015), and fully online matching is introduced by Huang et al. (JACM 2020). In this paper, we study the fractional versions of the two models. We improve three out of the four state-of-the-art upper and lower bounds of the two models. For fully online matching, we design a $0.6$-competitive algorithm and prove no algorithm can be $0.613$-competitive. For online matching with general vertex arrival, we prove no algorithm can be $0.584$-competitive. Moreover, we give an arguably more intuitive algorithm for the general vertex arrival model, compared to the algorithm of Wang and Wong, while attaining the same competitive ratio of $0.526$.

cs.DS↗

arXiv2021

Online Food Delivery to Minimize Maximum Flow Time

We study a common delivery problem encountered in nowadays online food-ordering platforms: Customers order dishes online, and the restaurant delivers the food after receiving the order. Specifically, we study a problem where $k$ vehicles of capacity $c$ are serving a set of requests ordering food from one restaurant. After a request arrives, it can be served by a vehicle moving from the restaurant to its delivery location. We are interested in serving all requests while minimizing the maximum flow-time, i.e., the maximum time length a customer waits to receive his/her food after submitting the order. We show that the problem is hard in both offline and online settings: There is a hardness of approximation of $Ω(n)$ for the offline problem, and a lower bound of $Ω(n)$ on the competitive ratio of any online algorithm, where $n$ is number of points in the metric. Our main result is an $O(1)$-competitive online algorithm for the uncapaciated (i.e, $c = \infty$) food delivery problem on tree metrics. Then we consider the speed-augmentation model. We develop an exponential time $(1+ε)$-speeding $O(1/ε)$-competitive algorithm for any $ε> 0$. A polynomial time algorithm can be obtained with a speeding factor of $α_{TSP}+ ε$ or $α_{CVRP}+ ε$, depending on whether the problem is uncapacitated. Here $α_{TSP}$ and $α_{CVRP}$ are the best approximation factors for the traveling salesman (TSP) and capacitated vehicle routing (CVRP) problems respectively. We complement the results with some negative ones.

cs.DS↗

arXiv2021

Control Barrier Function Meets Interval Analysis: Safety-Critical Control with Measurement and Actuation Uncertainties

This paper presents a framework for designing provably safe feedback controllers for sampled-data control affine systems with measurement and actuation uncertainties. Based on the interval Taylor model of nonlinear functions, a sampled-data control barrier function (CBF) condition is proposed which ensures the forward invariance of a safe set for sampled-data systems. Reachable set overapproximation and Lasserre's hierarchy of polynomial optimization are used for finding the margin term in the sampled-data CBF condition. Sufficient conditions for a safe controller in the presence of measurement and actuation uncertainties are proposed, for CBFs with relative degree 1 and higher relative degree individually. The effectiveness of the proposed method is illustrated by two numerical examples and an experimental example that implements the proposed controller on the Crazyflie quadcopter in real-time.

math.OC↗

arXiv2021

Certified Robustness to Programmable Transformations in LSTMs

Deep neural networks for natural language processing are fragile in the face of adversarial examples -- small input perturbations, like synonym substitution or word duplication, which cause a neural network to change its prediction. We present an approach to certifying the robustness of LSTMs (and extensions of LSTMs) and training models that can be efficiently certified. Our approach can certify robustness to intractably large perturbation spaces defined programmatically in a language of string transformations. Our evaluation shows that (1) our approach can train models that are more robust to combinations of string transformations than those produced using existing techniques; (2) our approach can show high certification accuracy of the resulting models.

cs.LG↗

arXiv2021

RadGraph: Extracting Clinical Entities and Relations from Radiology Reports

Extracting structured clinical information from free-text radiology reports can enable the use of radiology report information for a variety of critical healthcare applications. In our work, we present RadGraph, a dataset of entities and relations in full-text chest X-ray radiology reports based on a novel information extraction schema we designed to structure radiology reports. We release a development dataset, which contains board-certified radiologist annotations for 500 radiology reports from the MIMIC-CXR dataset (14,579 entities and 10,889 relations), and a test dataset, which contains two independent sets of board-certified radiologist annotations for 100 radiology reports split equally across the MIMIC-CXR and CheXpert datasets. Using these datasets, we train and test a deep learning model, RadGraph Benchmark, that achieves a micro F1 of 0.82 and 0.73 on relation extraction on the MIMIC-CXR and CheXpert test sets respectively. Additionally, we release an inference dataset, which contains annotations automatically generated by RadGraph Benchmark across 220,763 MIMIC-CXR reports (around 6 million entities and 4 million relations) and 500 CheXpert reports (13,783 entities and 9,908 relations) with mappings to associated chest radiographs. Our freely available dataset can facilitate a wide range of research in medical natural language processing, as well as computer vision and multi-modal learning when linked to chest radiographs.

cs.CL↗