arXiv subjects

Zili Zhou

Publications and source records attributed to Zili Zhou.

14 recordsLinked to original sources

arXiv2025

Structure-Aware Radar-Camera Depth Estimation

Radar has gained much attention in autonomous driving due to its accessibility and robustness. However, its standalone application for depth perception is constrained by issues of sparsity and noise. Radar-camera depth estimation offers a more promising complementary solution. Despite significant progress, current approaches fail to produce satisfactory dense depth maps, due to the unsatisfactory processing of the sparse and noisy radar data. They constrain the regions of interest for radar points in rigid rectangular regions, which may introduce unexpected errors and confusions. To address these issues, we develop a structure-aware strategy for radar depth enhancement, which provides more targeted regions of interest by leveraging the structural priors of RGB images. Furthermore, we design a Multi-Scale Structure Guided Network to enhance radar features and preserve detailed structures, achieving accurate and structure-detailed dense metric depth estimation. Building on these, we propose a structure-aware radar-camera depth estimation framework, named SA-RCD. Extensive experiments demonstrate that our SA-RCD achieves state-of-the-art performance on the nuScenes dataset. Our code will be available at https://github.com/FreyZhangYeh/SA-RCD.

cs.CV↗

arXiv2023

PhysNLU: A Language Resource for Evaluating Natural Language Understanding and Explanation Coherence in Physics

In order for language models to aid physics research, they must first encode representations of mathematical and natural language discourse which lead to coherent explanations, with correct ordering and relevance of statements. We present a collection of datasets developed to evaluate the performance of language models in this regard, which measure capabilities with respect to sentence ordering, position, section prediction, and discourse coherence. Analysis of the data reveals equations and sub-disciplines which are most common in physics discourse, as well as the sentence-level frequency of equations and expressions. We present baselines that demonstrate how contemporary language models are challenged by coherence related tasks in physics, even when trained on mathematical natural language objectives.

cs.CL↗

arXiv2022

COV19IR : COVID-19 Domain Literature Information Retrieval

Increasing number of COVID-19 research literatures cause new challenges in effective literature screening and COVID-19 domain knowledge aware Information Retrieval. To tackle the challenges, we demonstrate two tasks along withsolutions, COVID-19 literature retrieval, and question answering. COVID-19 literature retrieval task screens matching COVID-19 literature documents for textual user query, and COVID-19 question answering task predicts proper text fragments from text corpus as the answer of specific COVID-19 related questions. Based on transformer neural network, we provided solutions to implement the tasks on CORD-19 dataset, we display some examples to show the effectiveness of our proposed solutions.

cs.IR↗

arXiv2022

Conoscopic interferometry for optimal acoustic pulse detection in ultrafast acoustics

Conoscopic interferometry is a promising detection technique for ultrafast acoustics. By focusing a probe beam through a birefringent crystal before passing it through a polarizer, conoscopic interferences sculpt the spatial profile of the beam. The use of these patterns for acoustic wave detection revealed a higher detection sensitivity over existing techniques, such as reflectometry and beam distortion detection. However, the physical origin of the increased sensitivity is unknown. In this work, we present a model, describing the sensitivity behaviour of conoscopic interferometry with respect to the quarter-wave plate orientation and the diaphragm aperture, which is validated experimentally. Using the model, we optimize the detection sensitivity of conoscopic interferometry. We obtain a maximal sensitivity of detection when placing the diaphragm edge on the dark fringes of the conoscopic interference patterns. In the configurations studied in this work, conoscopic interferometry can be 8x more sensitive to acoustic waves than beam distortion detection.

physics.app-ph↗

arXiv2022

Transformers and the representation of biomedical background knowledge

Specialised transformers-based models (such as BioBERT and BioMegatron) are adapted for the biomedical domain based on publicly available biomedical corpora. As such, they have the potential to encode large-scale biological knowledge. We investigate the encoding and representation of biological knowledge in these models, and its potential utility to support inference in cancer precision medicine - namely, the interpretation of the clinical significance of genomic alterations. We compare the performance of different transformer baselines; we use probing to determine the consistency of encodings for distinct entities; and we use clustering methods to compare and contrast the internal properties of the embeddings for genes, variants, drugs and diseases. We show that these models do indeed encode biological knowledge, although some of this is lost in fine-tuning for specific tasks. Finally, we analyse how the models behave with regard to biases and imbalances in the dataset.

cs.CL↗

arXiv2022

Active entailment encoding for explanation tree construction using parsimonious generation of hard negatives

Entailment trees have been proposed to simulate the human reasoning process of explanation generation in the context of open--domain textual question answering. However, in practice, manually constructing these explanation trees proves a laborious process that requires active human involvement. Given the complexity of capturing the line of reasoning from question to the answer or from claim to premises, the issue arises of how to assist the user in efficiently constructing multi--level entailment trees given a large set of available facts. In this paper, we frame the construction of entailment trees as a sequence of active premise selection steps, i.e., for each intermediate node in an explanation tree, the expert needs to annotate positive and negative examples of premise facts from a large candidate list. We then iteratively fine--tune pre--trained Transformer models with the resulting positive and tightly controlled negative samples and aim to balance the encoding of semantic relationships and explanatory entailment relationships. Experimental evaluation confirms the measurable efficiency gains of the proposed active fine--tuning method in facilitating entailment trees construction: up to 20\% improvement in explanatory premise selection when compared against several alternatives.

cs.CL↗

arXiv2021

Encoding Explanatory Knowledge for Zero-shot Science Question Answering

This paper describes N-XKT (Neural encoding based on eXplanatory Knowledge Transfer), a novel method for the automatic transfer of explanatory knowledge through neural encoding mechanisms. We demonstrate that N-XKT is able to improve accuracy and generalization on science Question Answering (QA). Specifically, by leveraging facts from background explanatory knowledge corpora, the N-XKT model shows a clear improvement on zero-shot QA. Furthermore, we show that N-XKT can be fine-tuned on a target QA dataset, enabling faster convergence and more accurate results. A systematic analysis is conducted to quantitatively analyze the performance of the N-XKT model and the impact of different categories of knowledge on the zero-shot generalization task.

cs.CL↗

arXiv2020

On the correlation of earthquake occurrence among major fault zones in the eastern margin of the Tibetan Plateau by Big Data Analysis

The subsequent series of responses to big events may exhibit a synchronicity of event number, frequency and energy release in different fault zones. This synchronicity is a reliable source for probing non-intuitive geological structures, assessing regional seismicity hazard map and even predicting the next big events. The synchronicity of main faults in the eastern margin of the Qinghai-Tibetan Plateau is still unknown to us. We propose to examine the correlation of earthquake occurrence among different fault zones to indicate this synchronicity, and to obtain a preliminary understanding of geodynamics processes and the unrecognized characteristics of deep evolution in the eastern margin of the Qinghai-Tibetan Plateau. We estimate temporal changes of completeness level, frequency seismicity, and intensity seismicity, referring respectively to Mc, Z, and E values, of 21 main fault zones, using a seismic catalogue from 1970 to 2015. Our results reveal that six fault zone pairs of fault zones exhibit relative high correlation (>0.6) by all three indicators, while four fault zone pairs are non-adjacent with close internal affinity offsetting the limit of spatial distance, such as the pair of Rongjing-mabian fault and Minjiang-huya fault. Most strikingly, some fault zone pairs showing typical high correlation (>0.8) of seismicity frequency or seismicity intensity, the faults surprisingly belong to neither the same seismic belt nor the same geological block, exhibiting a regional scale remote triggering pattern of earthquakes or structures. An embryonic pattern to predict the next possible events will also be presented. This correlation analysis discovers a previously unrecognized strong coupling relationship among main faults with high earthquake risk in the eastern margin of the Qinghai-Tibetan Plateau.

physics.geo-ph↗

arXiv2020

Multi-Fusion Chinese WordNet (MCW) : Compound of Machine Learning and Manual Correction

Princeton WordNet (PWN) is a lexicon-semantic network based on cognitive linguistics, which promotes the development of natural language processing. Based on PWN, five Chinese wordnets have been developed to solve the problems of syntax and semantics. They include: Northeastern University Chinese WordNet (NEW), Sinica Bilingual Ontological WordNet (BOW), Southeast University Chinese WordNet (SEW), Taiwan University Chinese WordNet (CWN), Chinese Open WordNet (COW). By using them, we found that these word networks have low accuracy and coverage, and cannot completely portray the semantic network of PWN. So we decided to make a new Chinese wordnet called Multi-Fusion Chinese Wordnet (MCW) to make up those shortcomings. The key idea is to extend the SEW with the help of Oxford bilingual dictionary and Xinhua bilingual dictionary, and then correct it. More specifically, we used machine learning and manual adjustment in our corrections. Two standards were formulated to help our work. We conducted experiments on three tasks including relatedness calculation, word similarity and word sense disambiguation for the comparison of lemma's accuracy, at the same time, coverage also was compared. The results indicate that MCW can benefit from coverage and accuracy via our method. However, it still has room for improvement, especially with lemmas. In the future, we will continue to enhance the accuracy of MCW and expand the concepts in it.

cs.CL↗

arXiv2016

Experimental investigation of the detection mechanism in WSi nanowire superconducting single photon detectors

We use quantum detector tomography to investigate the detection mechanism in WSi nanowire superconducting single photon detectors (SSPDs). To this purpose, we fabricated a 250nm wide and 250nm long WSi nanowire and measured its response to impinging photons with wavelengths ranging from $λ$ = 900 nm to $λ$ = 1650 nm. Tomographic measurements show that the detector response depends on the total excitation energy only. Moreover, for energies Et > 0.8eV the current energy relation is linear, similar to what was observed in NbN nanowires, whereas the current-energy relation deviates from linear behaviour for total energies below 0.8eV.

cond-mat.mes-hall↗

arXiv2014

Inhomogeneous critical current in nanowire superconducting single-photon detectors

A superconducting thin film with uniform properties is the key to realize nanowire superconducting single-photon detectors (SSPDs) with high performance and high yield. To investigate the uniformity of NbN films, we introduce and characterize simple detectors consisting of short nanowires with length ranging from 100nm to 15μm. Our nanowires, contrary to meander SSPDs, allow probing the homogeneity of NbN at the nanoscale. Experimental results, endorsed by a microscopic model, show the strongly inhomogeneous nature of NbN films on the sub-100nm scale.

cond-mat.supr-con↗

arXiv2013

Superconducting series nanowire detector counting up to twelve photons

We demonstrate a superconducting photon-number-resolving detector capable of resolving up to twelve photons at telecommunication wavelengths. It is based on a series array of twelve superconducting NbN nanowire elements, each connected in parallel with an integrated resistor. The photon-induced voltage signals from the twelve elements are summed up into a single readout pulse with a height proportional to the detected photon number. Thirteen distinct output levels corresponding to the detection of n=0-12 photons are observed experimentally. A detailed analysis of the excess noise shows the potential of scaling to an even larger dynamic range.

quant-ph↗

arXiv2013

Universal Response Curve for Nanowire Superconducting Single-Photon Detectors

Using detector tomography, we investigate the detection mechanism in NbN-based superconducting single photon detectors (SSPDs). We demonstrate that the detection probability uniquely depends on a particular linear combination of bias current and energy, for a large variation of bias currents, input energies and detection probabilities, producing a universal detection curve. We obtain this result by studying multiphoton excitations in a nanodetector with a sparsity-based tomographic method that allows factoring out of the optical absorption. We discuss the implication of our model system for the understanding of meander-type SSPDs.

quant-ph↗

arXiv2012

Ultrasensitive N-photon interferometric autocorrelator

We demonstrate a novel method to measure the Nth-order (N=1, 2, 3, 4) interferometric autocorrelation with high sensitivity and temporal resolution. It is based on the combination of linear absorption and nonlinear detection in a superconducting nanodetector, providing much higher efficiency than methods based on all-optical nonlinearities. Its temporal resolution is only limited by the quasi-particle energy relaxation time, which is directly measured to be in the 20 ps range for the NbN films used in this work. We present a general model of interferometric autocorrelation with these nonlinear detectors and discuss the comparison with other approaches and possible improvements.

physics.optics↗