arXiv subjects

Yuzhe Zhang

Publications and source records attributed to Yuzhe Zhang.

47 records · Page 3Linked to original sources

arXiv2023

rPPG-Toolbox: Deep Remote PPG Toolbox

Camera-based physiological measurement is a fast growing field of computer vision. Remote photoplethysmography (rPPG) utilizes imaging devices (e.g., cameras) to measure the peripheral blood volume pulse (BVP) via photoplethysmography, and enables cardiac measurement via webcams and smartphones. However, the task is non-trivial with important pre-processing, modeling, and post-processing steps required to obtain state-of-the-art results. Replication of results and benchmarking of new models is critical for scientific progress; however, as with many other applications of deep learning, reliable codebases are not easy to find or use. We present a comprehensive toolbox, rPPG-Toolbox, that contains unsupervised and supervised rPPG models with support for public benchmark datasets, data augmentation, and systematic evaluation: \url{https://github.com/ubicomplab/rPPG-Toolbox}

cs.CV↗

arXiv2023

Training Robust Deep Physiological Measurement Models with Synthetic Video-based Data

Recent advances in supervised deep learning techniques have demonstrated the possibility to remotely measure human physiological vital signs (e.g., photoplethysmograph, heart rate) just from facial videos. However, the performance of these methods heavily relies on the availability and diversity of real labeled data. Yet, collecting large-scale real-world data with high-quality labels is typically challenging and resource intensive, which also raises privacy concerns when storing personal bio-metric data. Synthetic video-based datasets (e.g., SCAMPS \cite{mcduff2022scamps}) with photo-realistic synthesized avatars are introduced to alleviate the issues while providing high-quality synthetic data. However, there exists a significant gap between synthetic and real-world data, which hinders the generalization of neural models trained on these synthetic datasets. In this paper, we proposed several measures to add real-world noise to synthetic physiological signals and corresponding facial videos. We experimented with individual and combined augmentation methods and evaluated our framework on three public real-world datasets. Our results show that we were able to reduce the average MAE from 6.9 to 2.0.

cs.CV↗

arXiv2023

How to Rationally Select Your Delegatee in PoS

This paper centers around a simple yet crucial question for everyday users: How should one choose their delegated validators within proof-of-stake (PoS) protocols, particularly in the context of Ethereum 2.0? This has been a long-overlooked gap, as existing studies have primarily focused on inter-committee (validator set) behaviors and activities, while neglecting the dynamic formation of committees, especially for individual stakeholders seeking reliable validators. Our study bridges this gap by diving into the delegation process (normal users delegate their small-value tokens to delegatees who later act as validators) before entering an actual consensus phase. We propose a Bayesian model to quantify normal users' trust in delegatees, which we further incorporate into a game-theoretical model to simulate users' reactions against a set of critical factors identified through extensive research (including 10+ staking service provider as well as 30+ PoS blockchains). Our results reveal that users tend to choose their delegatees and utilize their tokens by carefully weighing the delegation cost, the behaviors of other users, and the reputation of delegatees, ultimately reaching a Nash equilibrium. Unfortunately, the collective trend significantly increases the likelihood of token concentration on a small number of delegatees.

cs.CR↗

arXiv2023

Fast shimming algorithm based on Bayesian optimization for magnetic resonance based dark matter search

The sensitivity and accessible mass range of magnetic resonance searches for axionlike dark matter depends on the homogeneity of applied magnetic fields. Optimizing homogeneity through shimming requires exploring a large parameter space which can be prohibitively time consuming. We have automated the process of tuning the shim-coil currents by employing an algorithm based on Bayesian optimization. This method is especially suited for applications where the duration of a single optimization step prohibits exploring the parameter space extensively or when there is no prior information on the optimal operation point. Using the Cosmic Axion Spin Precession Experiment (CASPEr)-gradient low-field apparatus, we show that for our setup this method converges after approximately 30 iterations to a sub-10 parts-per-million field homogeneity which is desirable for our dark matter search.

astro-ph.CO↗

arXiv2023

Frequency-scanning considerations in axionlike dark matter spin-precession experiments

Galactic dark matter may consist of axionlike particles (ALPs) that can be described as an "ultralight bosonic field" oscillating at the ALP Compton frequency. The ALP field can be searched for using nuclear magnetic resonance (NMR), where resonant precession of spins of a polarized sample can be sensitively detected. The ALP mass to which the experiment is sensitive is scanned by sweeping the bias magnetic field. The scanning either results in detection of ALP dark matter or rules out ALP dark matter with sufficiently strong couplings to nuclear spins over the range of ALP masses corresponding to the covered span of Larmor frequencies. In this work, scanning strategies are analyzed with the goal of optimizing the parameter-space coverage via a proper choice of experimental parameters (e.g., the effective transverse relaxation time).

physics.atom-ph↗

arXiv2022

RAPS: A Novel Few-Shot Relation Extraction Pipeline with Query-Information Guided Attention and Adaptive Prototype Fusion

Few-shot relation extraction (FSRE) aims at recognizing unseen relations by learning with merely a handful of annotated instances. To generalize to new relations more effectively, this paper proposes a novel pipeline for the FSRE task based on queRy-information guided Attention and adaptive Prototype fuSion, namely RAPS. Specifically, RAPS first derives the relation prototype by the query-information guided attention module, which exploits rich interactive information between the support instances and the query instances, in order to obtain more accurate initial prototype representations. Then RAPS elaborately combines the derived initial prototype with the relation information by the adaptive prototype fusion mechanism to get the integrated prototype for both train and prediction. Experiments on the benchmark dataset FewRel 1.0 show a significant improvement of our method against state-of-the-art methods.

cs.CL↗

arXiv2022

Low-rank Latent Matrix-factor Prediction Modeling for Generalized High-dimensional Matrix-variate Regression

Motivated by diagnosing the COVID-19 disease using 2D image biomarkers from computed tomography (CT) scans, we propose a novel latent matrix-factor regression model to predict responses that may come from an exponential distribution family, where covariates include high-dimensional matrix-variate biomarkers. A latent generalized matrix regression (LaGMaR) is formulated, where the latent predictor is a low-dimensional matrix factor score extracted from the low-rank signal of the matrix variate through a cutting-edge matrix factor model. Unlike the general spirit of penalizing vectorization plus the necessity of tuning parameters in the literature, instead, our prediction modeling in LaGMaR conducts dimension reduction that respects the geometry characteristic of intrinsic two-dimensional structure of the matrix covariate and thus avoids iteration. This greatly relieves the computation burden, and meanwhile maintains structural information so that the latent matrix factor feature can perfectly replace the intractable matrix-variate owing to high-dimensionality. The estimation procedure of LaGMaR is subtly derived by transforming the bilinear form matrix factor model onto a high-dimensional vector factor model, so that the method of principle components can be applied. We establish bilinear-form consistency of the estimated matrix coefficient of the latent predictor and consistency of prediction. The proposed approach can be implemented conveniently. Through simulation experiments, prediction capability of LaGMaR is shown to outperform existing penalized methods under diverse scenarios of generalized matrix regressions. Through the application to a real COVID-19 dataset, the proposed approach is shown to predict efficiently the COVID-19.

stat.AP↗

arXiv2022

FinBERT-MRC: financial named entity recognition using BERT under the machine reading comprehension paradigm

Financial named entity recognition (FinNER) from literature is a challenging task in the field of financial text information extraction, which aims to extract a large amount of financial knowledge from unstructured texts. It is widely accepted to use sequence tagging frameworks to implement FinNER tasks. However, such sequence tagging models cannot fully take advantage of the semantic information in the texts. Instead, we formulate the FinNER task as a machine reading comprehension (MRC) problem and propose a new model termed FinBERT-MRC. This formulation introduces significant prior information by utilizing well-designed queries, and extracts start index and end index of target entities without decoding modules such as conditional random fields (CRF). We conduct experiments on a publicly available Chinese financial dataset ChFinAnn and a real-word bussiness dataset AdminPunish. FinBERT-MRC model achieves average F1 scores of 92.78% and 96.80% on the two datasets, respectively, with average F1 gains +3.94% and +0.89% over some sequence tagging models including BiLSTM-CRF, BERT-Tagger, and BERT-CRF. The source code is available at https://github.com/zyz0000/FinBERT-MRC.

cs.CL↗

arXiv2021

Tracking Truth by Weighting Proxies in Liquid Democracy

We study wisdom of the crowd effects in liquid democracy when agents are allowed to apportion weights to proxies by mixing their delegations. We show that in this setting -- unlike in the standard one where votes are always delegated in full to one proxy -- it becomes possible to identify delegation structures that optimize the truth-tracking accuracy of the group. We contrast this centralized solution with the group accuracy obtained in equilibrium when agents interact by greedily trying to maximize their own individual accuracy through mixed delegations, and study the price of anarchy of these games. While equilibria with mixed delegations may be as bad as in the standard delegations setting, they are never worse and may sometimes be better.

cs.GT↗

arXiv2020

Summarizing Unstructured Logs in Online Services

Logs are one of the most valuable data sources for managing large-scale online services. After a failure is detected/diagnosed/predicted, operators still have to inspect the raw logs to gain a summarized view before take actions. However, manual or rule-based log summarization has become inefficient and ineffective. In this work, we propose LogSummary, an automatic, unsupervised end-to-end log summarization framework for online services. LogSummary obtains the summarized triples of important logs for a given log sequence. It integrates a novel information extraction method taking both semantic information and domain knowledge into consideration, with a new triple ranking approach using the global knowledge learned from all logs. Given the lack of a publicly-available gold standard for log summarization, we have manually labelled the summaries of four open-source log datasets and made them publicly available. The evaluation on these datasets as well as the case studies on real-world logs demonstrate that LogSummary produces a highly representative (average ROUGE F1 score of 0.741) summaries. We have packaged LogSummary into an open-source toolkit and hope that it can benefit for future NLP-powered summarization works.

cs.SE↗

arXiv2020

Power in Liquid Democracy

The paper develops a theory of power for delegable proxy voting systems. We define a power index able to measure the influence of both voters and delegators. Using this index, which we characterize axiomatically, we extend an earlier game-theoretic model by incorporating power-seeking behavior by agents. We analytically study the existence of pure strategy Nash equilibria in such a model. Finally, by means of simulations, we study the effect of relevant parameters on the emergence of power inequalities in the model.

cs.GT↗