arXiv subjects

Yucheng Hu

Publications and source records attributed to Yucheng Hu.

35 records · Page 2Linked to original sources

arXiv2025

The effect of a band gap gradient on the radiative losses in the open circuit voltage of solar cells

The radiative open circuit voltage loss in a solar cell occurs because the absorptance spectrum near the band gap shows gradual increase rather than sharp step function like transition. This broadening effect has been attributed to band gap fluctuations and or to Urbach tails. In this report, we use modelling based on Planck s generalized law to distinguish between these two effects. Our results demonstrate that Urbach tails have only a minimal effect on the absorptance edge broadening and clarify that even an ideal direct semiconductor with no band gap fluctuations shows broadening at the absorptance onset. Furthermore, state of the art inorganic thin film solar cells often incorporate a band gap gradient across their thickness, which can further contribute to absorptance broadening. Using Cu(In,Ga)Se2 (CIGSe) absorbers as a case study, we perform a comprehensive analysis of voltage losses through absolute photoluminescence and electroluminescence spectroscopy, combined with photospectrometry and high-spatial-resolution cathodoluminescence measurements. We find that the loss analysis based on the combination of radiative, generation and non-radiative losses is complete. Samples with a graded band gap profile show more pronounced broadening of the absorptance onset and up to 16 mV higher radiative losses compared to the samples with uniform band gap. There is indication, that band gap-graded samples also have larger lateral band gap inhomogeneity.

cond-mat.mtrl-sci↗

arXiv2025

HiRT: Enhancing Robotic Control with Hierarchical Robot Transformers

Large Vision-Language-Action (VLA) models, leveraging powerful pre trained Vision-Language Models (VLMs) backends, have shown promise in robotic control due to their impressive generalization ability. However, the success comes at a cost. Their reliance on VLM backends with billions of parameters leads to high computational costs and inference latency, limiting the testing scenarios to mainly quasi-static tasks and hindering performance in dynamic tasks requiring rapid interactions. To address these limitations, this paper proposes HiRT, a Hierarchical Robot Transformer framework that enables flexible frequency and performance trade-off. HiRT keeps VLMs running at low frequencies to capture temporarily invariant features while enabling real-time interaction through a high-frequency vision-based policy guided by the slowly updated features. Experiment results in both simulation and real-world settings demonstrate significant improvements over baseline methods. Empirically, in static tasks, we double the control frequency and achieve comparable success rates. Additionally, on novel real-world dynamic ma nipulation tasks which are challenging for previous VLA models, HiRT improves the success rate from 48% to 75%.

cs.CV↗

arXiv2025

Improving Vision-Language-Action Model with Online Reinforcement Learning

Recent studies have successfully integrated large vision-language models (VLMs) into low-level robotic control by supervised fine-tuning (SFT) with expert robotic datasets, resulting in what we term vision-language-action (VLA) models. Although the VLA models are powerful, how to improve these large models during interaction with environments remains an open question. In this paper, we explore how to further improve these VLA models via Reinforcement Learning (RL), a commonly used fine-tuning technique for large models. However, we find that directly applying online RL to large VLA models presents significant challenges, including training instability that severely impacts the performance of large models, and computing burdens that exceed the capabilities of most local machines. To address these challenges, we propose iRe-VLA framework, which iterates between Reinforcement Learning and Supervised Learning to effectively improve VLA models, leveraging the exploratory benefits of RL while maintaining the stability of supervised learning. Experiments in two simulated benchmarks and a real-world manipulation suite validate the effectiveness of our method.

cs.RO↗

arXiv2024

Prediction with Action: Visual Policy Learning via Joint Denoising Process

Diffusion models have demonstrated remarkable capabilities in image generation tasks, including image editing and video creation, representing a good understanding of the physical world. On the other line, diffusion models have also shown promise in robotic control tasks by denoising actions, known as diffusion policy. Although the diffusion generative model and diffusion policy exhibit distinct capabilities--image prediction and robotic action, respectively--they technically follow a similar denoising process. In robotic tasks, the ability to predict future images and generate actions is highly correlated since they share the same underlying dynamics of the physical world. Building on this insight, we introduce PAD, a novel visual policy learning framework that unifies image Prediction and robot Action within a joint Denoising process. Specifically, PAD utilizes Diffusion Transformers (DiT) to seamlessly integrate images and robot states, enabling the simultaneous prediction of future images and robot actions. Additionally, PAD supports co-training on both robotic demonstrations and large-scale video datasets and can be easily extended to other robotic modalities, such as depth images. PAD outperforms previous methods, achieving a significant 26.3% relative improvement on the full Metaworld benchmark, by utilizing a single text-conditioned visual policy within a data-efficient imitation learning setting. Furthermore, PAD demonstrates superior generalization to unseen tasks in real-world robot manipulation settings with 28.0% success rate increase compared to the strongest baseline. Project page at https://sites.google.com/view/pad-paper

cs.RO↗

arXiv2024

Improved sequentially processed Cu(In,Ga)(S,Se)2 by Ag alloying

Alloying small quantities of silver into Cu(In,Ga)Se2 was shown to improve the efficiency for wide and low band gap solar cells. We study low band gap industrial Cu(In,Ga)(S,Se)2 absorbers, substituting less than 10% of the copper with silver, using absolute photoluminescence and cathodoluminescence spectroscopy. Silver improves the grain size and promotes the interdiffusion of Ga and In across the depth of the absorber, resulting in a smoother band gap gradient. However, a certain lateral inhomogeneity is observed near the front and back sides. The non-radiative losses in the bare absorbers are reduced by up to 30 meV.

cond-mat.mtrl-sci↗

arXiv2021

Object Detection Based Handwriting Localization

We present an object detection based approach to localize handwritten regions from documents, which initially aims to enhance the anonymization during the data transmission. The concatenated fusion of original and preprocessed images containing both printed texts and handwritten notes or signatures are fed into the convolutional neural network, where the bounding boxes are learned to detect the handwriting. Afterwards, the handwritten regions can be processed (e.g. replaced with redacted signatures) to conceal the personally identifiable information (PII). This processing pipeline based on the deep learning network Cascade R-CNN works at 10 fps on a GPU during the inference, which ensures the enhanced anonymization with minimal computational overheads. Furthermore, the impressive generalizability has been empirically showcased: the trained model based on the English-dominant dataset works well on the fictitious unseen invoices, even in Chinese. The proposed approach is also expected to facilitate other tasks such as handwriting recognition and signature verification.

cs.CV↗

arXiv2021

Solution Landscapes of the Simplified Ericksen--Leslie Model and its Comparison with the Reduced Landau--de Gennes Model

We investigate the solution landscapes of a simplified Ericksen--Leslie (sEL) vector model for nematic liquid crystals, confined in a two-dimensional square domain with tangent boundary conditions. An efficient numerical algorithm is developed to construct the solution landscapes by utilizing the symmetry properties of the model and the domain. Since the sEL model and the reduced Landau--de Gennes (rLdG) models can be viewed as Ginzburg--Landau functionals, we systematically compute the solution landscapes of the sEL model, for different domain sizes, and compare with the solution landscapes of the corresponding rLdG models. There are many similarities, including the stable diagonal and rotated states, bifurcation behaviors, and sub-solution landscapes with low-index saddle solutions. Significant disparities also exist between the two models. The sEL vector model exhibits the stable solution $C\pm$ with interior defects, high-index "fake defects" solutions, novel tessellating solutions, and certain types of distinctive dynamical pathways. The solution landscape approach provides a comprehensive and efficient way for model comparison and is applicable to a wide range of mathematical models in physics.

cond-mat.soft↗

arXiv2021

XCrossNet: Feature Structure-Oriented Learning for Click-Through Rate Prediction

Click-Through Rate (CTR) prediction is a core task in nowadays commercial recommender systems. Feature crossing, as the mainline of research on CTR prediction, has shown a promising way to enhance predictive performance. Even though various models are able to learn feature interactions without manual feature engineering, they rarely attempt to individually learn representations for different feature structures. In particular, they mainly focus on the modeling of cross sparse features but neglect to specifically represent cross dense features. Motivated by this, we propose a novel Extreme Cross Network, abbreviated XCrossNet, which aims at learning dense and sparse feature interactions in an explicit manner. XCrossNet as a feature structure-oriented model leads to a more expressive representation and a more precise CTR prediction, which is not only explicit and interpretable, but also time-efficient and easy to implement. Experimental studies on Criteo Kaggle dataset show significant improvement of XCrossNet over state-of-the-art models on both effectiveness and efficiency.

cs.IR↗

arXiv2019

Conservation-Dissipation Formalism for Soft Matter Physics: I. Equivalence with Doi's Variational Approach

In this paper, we proved that by choosing the proper variational function and variables, the variational approach proposed by M. Doi in soft matter physics was equivalent to the Conservation-Dissipation Formalism. To illustrate the correspondence between these two theories, several novel examples in soft matter physics, including particle diffusion in dilute solutions, polymer phase separation dynamics and nematic liquid crystal flows, were carefully examined. Based on our work, a deep connection among the generalized Gibbs relation, the second law of thermodynamics and the variational principle in non-equilibrium thermodynamics was revealed.

cond-mat.stat-mech↗

arXiv2019

Conservation-Dissipation Formalism for Soft Matter Physics: II. Application to Non-isothermal Nematic Liquid Crystals

To most existing non-equilibrium theories, the modeling of non-isothermal processes was a hard task. Intrinsic difficulties involved the non-equilibrium temperature, the coexistence of conserved energy and dissipative entropy, etc. In this paper, by taking the non-isothermal flow of nematic liquid crystals as a typical example, we illustrated that thermodynamically consistent models in either vectorial or tensorial forms could be constructed within the framework of Conservation-Dissipation Formalism (CDF). And the classical isothermal Ericksen-Leslie model and Qian-Sheng model were shown to be special cases of our new vectorial and tensorial models in the isothermal, incompressible and stationary limit. Most importantly, from above examples, it was learnt that mathematical modeling based on CDF could easily solve the issues relating with non-isothermal situations in a systematic way. The first and second laws of thermodynamics were satisfied simultaneously. The non-equilibrium temperature was defined self-consistently through the partial derivative of entropy function. Relaxation-type constitutive relations were constructed, which gave rise to the classical linear constitutive relations, like Newton's law and Fourier's law, in stationary limits. Therefore, CDF was expected to have a broad scope of applications in soft matter physics, especially under the complicated situations, such as non-isothermal, compressible and nanoscale systems.

cond-mat.soft↗

arXiv2018

Phase Transition Dynamics and Stochastic Resonance in Topologically Confined Nematic Liquid Crystals

Topological defects resulted from boundary constraints in confined liquid crystals have attracted extensive research interests. In this paper, we use numerical simulation to study the phase transition dynamics in the context of stochastic resonance in a bistable liquid crystal device containing defects. This device is made of nematic liquid crystals confined in a shallow square well, and is described by the planar Lebwohl-Lasher model. The stochastic phase transition processes of the system in the presence of a weak oscillating potential is simulated using an over-damped Langevin dynamics. Our simulation results reveal that, depending on system size, the phase transition may follow two distinct pathways: in small systems the pre-existing defect structures at the corners hold until the last stage and there is no newly formed defect point in the bulk during the phase transition, In large systems new defect points appear spontaneously in the bulk and eventually merge with the pre-existing defects at the corners. For both transition pathways stochastic resonance can be observed, but show dramatic difference in their responses to the boundary anchoring strength. In small systems we observe a "sticky-boundary" effect for a certain range of anchoring strength in which the phase transition gets stuck and stochastic resonance becomes de-activated. Our work demonstrates the dynamical interplay among defects, noises, and boundary conditions in confined liquid crystals.

cond-mat.soft↗

arXiv2018

A Deep Learning Approach to the Citywide Traffic Accident Risk Prediction

With the rapid development of urbanization, the boom of vehicle numbers has resulted in serious traffic accidents, which led to casualties and huge economic losses. The ability to predict the risk of traffic accident is important in the prevention of the occurrence of accidents and to reduce the damages caused by accidents in a proactive way. However, traffic accident risk prediction with high spatiotemporal resolution is difficult, mainly due to the complex traffic environment, human behavior, and lack of real-time traffic-related data. In this study, we collected big traffic accident data. By analyzing the spatial and temporal patterns of traffic accident frequency, we presented the spatiotemporal correlation of traffic accidents. Based on the patterns we found in analysis, we proposed a high accurate deep learning model based on recurrent neural network toward the prediction of traffic accident risk. The predictive accident risk can be potential applied to the traffic accident warning system. The proposed method can be integrated into an intelligent traffic control system toward a more reasonable traffic prediction and command organization.

cs.CY↗

arXiv2016

City traffic forecasting using taxi GPS data: A coarse-grained cellular automata model

City traffic is a dynamic system of enormous complexity. Modeling and predicting city traffic flow remains to be a challenge task and the main difficulties are how to specify the supply and demands and how to parameterize the model. In this paper we attempt to solve these problems with the help of large amount of floating car data. We propose a coarse-grained cellular automata model that simulates vehicles moving on uniform grids whose size are much larger compared with the microscopic cellular automata model. The car-car interaction in the microscopic model is replaced by the coupling between vehicles and coarse-grained state variables in our model. To parameterize the model, flux-occupancy relations are fitted from the historical data at every grids, which serve as the coarse-grained fundamental diagrams coupling the occupancy and speed. To evaluate the model, we feed it with the historical travel demands and trajectories obtained from the floating car data and use the model to predict road speed one hour into the future. Numerical results show that our model can capture the traffic flow pattern of the entire city and make reasonable predictions. The current work can be considered a prototype for a model-based forecasting system for city traffic.

cs.OH↗

arXiv2015

Collective Properties of a Transcription Initiation Model under Varying Environment

The dynamics of gene transcription is tightly regulated in eukaryotes. Recent experiments have revealed various kinds of transcriptional dynamics, such as RNA polymerase II pausing, that involves regulation at the transcription initiation stage, and the choice of different regulation pattern is closely related to the physiological functions of the target gene. Here we consider a simplified model of transcription initiation, a process including the assembly of transcription complex and the pausing and releasing of the RNA polymerase II. Focusing on the collective behaviors on a population level, we explore potential regulatory functions this model can offer. These functions include fast and synchronized response to environmental change, or long-term memory about the transcriptional status. As a proof of concept we also show that, by selecting different control mechanisms cells can adapt to different environments. These findings may help us better understand the design principles of transcriptional regulation.

q-bio.CB↗

arXiv2014

On the Disclination Lines of Nematic Liquid Crystals

Defects in liquid crystals are of great practical importance and theoretical interest. Despite tremendous efforts, predicting the location and transition of defects under various topological constraint and external field remains to be a challenge. We investigate defect patterns of nematic liquid crystals confined in three-dimensional spherical droplet and two-dimensional disk under different boundary conditions, within the Landau-de Gennes model. We implement a spectral method that numerically solves the Landau-de Gennes model with high accuracy, which allows us to study the detailed static structure of defects. We observe five types of defect structures. Among them the 1/2-disclination lines are the most stable structure at low temperature. Inspired by numerical results, we obtain the profile of disclination lines analytically. Moreover, the connection and difference between defect patterns under the Landau-de Gennes model and the Oseen-Frank model are discussed. Finally, four conjectures are made to summarize some important characteristics of defects in the Landau-de Gennes theory. This work is a continuing effort to deepen our understanding on defect patterns in nematic liquid crystals.

cond-mat.soft↗

arXiv2013

Cell Growth and Size Homeostasis in Silico

Cell growth in size is a complex process coordinated by intrinsic and environmental signals. In a recent work [Tzur et al., Science, 2009, 325:167-171], size distributions in an exponentially growing population of mammalian cells were used to infer the growth rate in size. The results suggest that cell growth is neither linear nor exponential, but subject to size-dependent regulation. To explain their data, we build a model in which the cell growth rate is controlled by the relative amount of mRNA and ribosomes in a cell. Plus a stochastic division rule, the evolutionary process of a population of cells can be simulated and the statistics of the in-silico population agree well with the experimental data. To further explore the model space, alternative growth models and division rules are studied. This work may serve as a starting point for us to understand the rational behind cell growth and size regulation using predictive models.

q-bio.CB↗

arXiv2006

On the Poisson Approximation to Photon Distribution for Faint Lasers

It is proved, that for a certain kind of input distribution, the strongly binomially attenuated photon number distribution can well be approximated by a Poisson distribution. This explains why we can adopt poissonian distribution as the photon number statistics for faint lasers. The error of such an approximation is quantitatively estimated. Numerical tests are carried out, which coincide with our theoretical estimations. This work lays a sound mathematical foundation for the well-known intuitive idea which has been widely used in quantum cryptography.

math-ph↗