arXiv subjects

Yuanbo Chen

Publications and source records attributed to Yuanbo Chen.

At least 19 recordsLinked to original sources

arXiv2026

Physics-Informed WiFi Sensing for Robust 3D Human Pose Estimation in Mobile and Cross-Environment Settings

Device-free human pose estimation using commodity WiFi signals has emerged as a promising paradigm for pervasive sensing in mobile computing systems. However, existing approaches often suffer from severe performance degradation when deployed across heterogeneous environments due to complex multipath propagation and domain shifts in wireless signals. In this paper, we present a physics-informed WiFi sensing framework for robust 3D human pose estimation under mobile and cross-environment settings. Our approach explicitly models wireless signal propagation characteristics and incorporates multipath-aware attention to capture environment-dependent signal variations. To further improve generalization, we introduce a disentangled representation learning scheme that separates pose-related features from environment-specific factors, enabling effective cross-domain adaptation without requiring extensive retraining. We implement our system using commodity WiFi devices and evaluate it on multiple public benchmarks, including Person-in-WiFi-3D and MM-Fi, as well as real-world deployments across diverse indoor environments. Experimental results demonstrate that our framework significantly improves robustness and generalization performance compared to state-of-the-art methods, particularly under cross-environment scenarios. These results highlight the potential of physics-informed wireless sensing for enabling reliable, scalable, and infrastructure-free human-centric applications in mobile computing systems.

eess.SP↗

arXiv2026

Bridging the Semantic-Collaborative Gap: An Asymmetric Graph Architecture for Cold-Start Item Recommendation

Collaborative filtering and graph-based recommendation models are highly effective because they leverage observed user interactions, but this dependence creates a fundamental cold-start challenge when newly added content has no interaction history. In Tubi's production retrieval system, this challenge is further constrained by the serving interface: new content must be assigned a standalone embedding immediately, and the model must also produce device embeddings suitable for approximate nearest-neighbor retrieval. We address this setting by formulating cold-start recommendation as an inductive graph-completion problem on a temporal bipartite device-content graph. We propose Shallow-RHS, an asymmetric link-prediction architecture in which the left-hand side (LHS) device tower leverages temporally valid watch-history message passing to capture collaborative signals, while the right-hand side (RHS) content tower is intentionally shallow with respect to the graph and encodes content solely from intrinsic features. The RHS tower does not use ID-based embeddings, content-side subgraphs, neighbor aggregation, or interaction-derived representations, forcing the content encoder to map intrinsic features into a collaborative-filtering-aware embedding space. After training, the learned content encoder generates embeddings for both warm and newly ingested content, enabling implicit graph completion through retrieval of warm surrogate neighbors. We further extend the same representation-completion principle to device cold-start by constructing cohort-based embeddings from demographic features. Large-scale online experiments demonstrate consistent relative improvements in content cold-start engagement, promotion speed, impression acquisition, and device cold-start engagement.

cs.IR↗

arXiv2024

Optimal Energy Dispatch of Grid-Connected Electric Vehicle Considering Lithium Battery Electrochemical Model

The grid-connected electric vehicles (EVs) serve as a promising regulating resource in the distribution grid with Vehicle-to-Grid (V2G) facilities. In the day-ahead stage, electric vehicle batteries (EVBs) need to be precisely dispatched and controlled to ensure high efficiency and prevent degradation. This article focuses on considering a refined battery model, i.e. the electrochemical model (EM), in the optimal dispatch of the local energy system with high penetration of EVs which replenish energy through V2G-equipped charge station and battery swapping station (BSS). In this paper, to utilize the EM efficiently, recursive EVB constraints and a corresponding matrix-based state update method are proposed based on EM power characterization. The charging EV state distribution is profiled and a multi-layer BSS model along with binary aggregation is proposed, in order to overcome the computation complexity of combining the refined battery constraints with the mixed integer optimization. Finally, a local energy system scenario is investigated for evaluation. The efficiency and effectiveness of EM consideration are assessed from the perspective of both the system and battery.

eess.SY↗

arXiv2024

AI-Driven Stylization of 3D Environments

In this system, we discuss methods to stylize a scene of 3D primitive objects into a higher fidelity 3D scene using novel 3D representations like NeRFs and 3D Gaussian Splatting. Our approach leverages existing image stylization systems and image-to-3D generative models to create a pipeline that iteratively stylizes and composites 3D objects into scenes. We show our results on adding generated objects into a scene and discuss limitations.

cs.CV↗

arXiv2024

Scalable Indoor Novel-View Synthesis using Drone-Captured 360 Imagery with 3D Gaussian Splatting

Scene reconstruction and novel-view synthesis for large, complex, multi-story, indoor scenes is a challenging and time-consuming task. Prior methods have utilized drones for data capture and radiance fields for scene reconstruction, both of which present certain challenges. First, in order to capture diverse viewpoints with the drone's front-facing camera, some approaches fly the drone in an unstable zig-zag fashion, which hinders drone-piloting and generates motion blur in the captured data. Secondly, most radiance field methods do not easily scale to arbitrarily large number of images. This paper proposes an efficient and scalable pipeline for indoor novel-view synthesis from drone-captured 360 videos using 3D Gaussian Splatting. 360 cameras capture a wide set of viewpoints, allowing for comprehensive scene capture under a simple straightforward drone trajectory. To scale our method to large scenes, we devise a divide-and-conquer strategy to automatically split the scene into smaller blocks that can be reconstructed individually and in parallel. We also propose a coarse-to-fine alignment strategy to seamlessly match these blocks together to compose the entire scene. Our experiments demonstrate marked improvement in both reconstruction quality, i.e. PSNR and SSIM, and computation time compared to prior approaches.

cs.CV↗

arXiv2024

Group Benefits Instances Selection for Data Purification

Manually annotating datasets for training deep models is very labor-intensive and time-consuming. To overcome such inferiority, directly leveraging web images to conduct training data becomes a natural choice. Nevertheless, the presence of label noise in web data usually degrades the model performance. Existing methods for combating label noise are typically designed and tested on synthetic noisy datasets. However, they tend to fail to achieve satisfying results on real-world noisy datasets. To this end, we propose a method named GRIP to alleviate the noisy label problem for both synthetic and real-world datasets. Specifically, GRIP utilizes a group regularization strategy that estimates class soft labels to improve noise robustness. Soft label supervision reduces overfitting on noisy labels and learns inter-class similarities to benefit classification. Furthermore, an instance purification operation globally identifies noisy labels by measuring the difference between each training sample and its class soft label. Through operations at both group and instance levels, our approach integrates the advantages of noise-robust and noise-cleaning methods and remarkably alleviates the performance degradation caused by noisy labels. Comprehensive experimental results on synthetic and real-world datasets demonstrate the superiority of GRIP over the existing state-of-the-art methods.

cs.LG↗

arXiv2023

Model-based Constrained MDP for Budget Allocation in Sequential Incentive Marketing

Sequential incentive marketing is an important approach for online businesses to acquire customers, increase loyalty and boost sales. How to effectively allocate the incentives so as to maximize the return (e.g., business objectives) under the budget constraint, however, is less studied in the literature. This problem is technically challenging due to the facts that 1) the allocation strategy has to be learned using historically logged data, which is counterfactual in nature, and 2) both the optimality and feasibility (i.e., that cost cannot exceed budget) needs to be assessed before being deployed to online systems. In this paper, we formulate the problem as a constrained Markov decision process (CMDP). To solve the CMDP problem with logged counterfactual data, we propose an efficient learning algorithm which combines bisection search and model-based planning. First, the CMDP is converted into its dual using Lagrangian relaxation, which is proved to be monotonic with respect to the dual variable. Furthermore, we show that the dual problem can be solved by policy learning, with the optimal dual variable being found efficiently via bisection search (i.e., by taking advantage of the monotonicity). Lastly, we show that model-based planing can be used to effectively accelerate the joint optimization process without retraining the policy for every dual variable. Empirical results on synthetic and real marketing datasets confirm the effectiveness of our methods.

cs.AI↗

arXiv2022

Multi-Scale User Behavior Network for Entire Space Multi-Task Learning

Modelling the user's multiple behaviors is an essential part of modern e-commerce, whose widely adopted application is to jointly optimize click-through rate (CTR) and conversion rate (CVR) predictions. Most of existing methods overlook the effect of two key characteristics of the user's behaviors: for each item list, (i) contextual dependence refers to that the user's behaviors on any item are not purely determinated by the item itself but also are influenced by the user's previous behaviors (e.g., clicks, purchases) on other items in the same sequence; (ii) multiple time scales means that users are likely to click frequently but purchase periodically. To this end, we develop a new multi-scale user behavior network named Hierarchical rEcurrent Ranking On the Entire Space (HEROES) which incorporates the contextual information to estimate the user multiple behaviors in a multi-scale fashion. Concretely, we introduce a hierarchical framework, where the lower layer models the user's engagement behaviors while the upper layer estimates the user's satisfaction behaviors. The proposed architecture can automatically learn a suitable time scale for each layer to capture the dynamic user's behavioral patterns. Besides the architecture, we also introduce the Hawkes process to form a novel recurrent unit which can not only encode the items' features in the context but also formulate the excitation or discouragement from the user's previous behaviors. We further show that HEROES can be extended to build unbiased ranking systems through combinations with the survival analysis technique. Extensive experiments over three large-scale industrial datasets demonstrate the superiority of our model compared with the state-of-the-art methods.

cs.IR↗

arXiv2022

Electrochemical Parameter Identification for Lithium-ion Battery Sources in Self-Sustained Transportation Energy Systems

Lithium-ion battery (LIB) sources have played an essential role in self-sustained transportation energy systems and have been widely deployed in the last few years. To realize reliable battery maintenance, identifying its electrochemical parameters is necessary. However, the battery model contains many parameters while the measurable states are only the current and voltage, inducing the identification inherently an ill-conditioned problem. A parameter identification approach is proposed, including the experiment, model, and algorithm. Electrochemical parameters are first grouped manually based on the physical properties and assigned to two sequenced tests for identification. The two tests named the quasi-static test and the dynamic test, are compressed on time for practical implementation. Proper optimization models and a sensitivity-oriented stepwise (SSO) optimization algorithm are developed to search for the optimal parameters efficiently. Typically, the Sobol method is applied to conduct the sensitivity analysis. Based on the sensitivity indexes, the SSO algorithm can decouple the mixed impacts of different parameters during the identification. For validation, numerical experiments on a typical NCM811 battery at different life stages are conducted. The proposed approach saves about half the time finding the proper parameter value. The identification accuracy of crucial parameters related to battery degradation can exceed 95\%. Case study results indicate that the identified parameters can not only improve the accuracy of the battery model but also be used as the indicator of the battery SOH.

eess.SY↗

arXiv2022

Quantum Walks with Indefinite Causal Order

In all existing quantum walk models, the assumption about a pre-existing fixed background causal structure is always made and has been taken for granted. Nevertheless, in this work we will get rid of this tacit assumption especially by introducing indefinite causal order ways of coin tossing and investigate this modern scenario. We find that an ideal-shape and fast-spreading uniform distribution can be prepared with our new model. First we will show how an always-symmetrical instantaneous distribution appears in an indefinite causal order quantum walk, which then paves the way for deriving conditions that enables one to interpret an evolved state as a superposition of its definite causal order counterparts which is however in general prohibited. This property of temporal superposition of states can be generalized from two-process scenarios to cases where arbitrary many $\mathcal{N}$ processes are involved. Finally, we demonstrate how a genuine uniform distribution emerges from an indefinite causal order quantum walk. Remarkably, besides the ideal shape of the distribution, our protocol has another powerful advantage, that is the speed of spatial spreading reaches exactly the theoretical limit in contrast to conventional cases where one may encounter the well-known issue of a $1/\sqrt{2}$ degeneration.

quant-ph↗

arXiv2022

Who to Watch Next: Two-side Interactive Networks for Live Broadcast Recommendation

With the prevalence of live broadcast business nowadays, a new type of recommendation service, called live broadcast recommendation, is widely used in many mobile e-commerce Apps. Different from classical item recommendation, live broadcast recommendation is to automatically recommend user anchors instead of items considering the interactions among triple-objects (i.e., users, anchors, items) rather than binary interactions between users and items. Existing methods based on binary objects, ranging from early matrix factorization to recently emerged deep learning, obtain objects' embeddings by mapping from pre-existing features. Directly applying these techniques would lead to limited performance, as they are failing to encode collaborative signals among triple-objects. In this paper, we propose a novel TWo-side Interactive NetworkS (TWINS) for live broadcast recommendation. In order to fully use both static and dynamic information on user and anchor sides, we combine a product-based neural network with a recurrent neural network to learn the embedding of each object. In addition, instead of directly measuring the similarity, TWINS effectively injects the collaborative effects into the embedding process in an explicit manner by modeling interactive patterns between the user's browsing history and the anchor's broadcast history in both item and anchor aspects. Furthermore, we design a novel co-retrieval technique to select key items among massive historic records efficiently. Offline experiments on real large-scale data show the superior performance of the proposed TWINS, compared to representative methods; and further results of online experiments on Diantao App show that TWINS gains average performance improvement of around 8% on ACTR metric, 3% on UCTR metric, 3.5% on UCVR metric.

cs.IR↗

arXiv2021

Fundamental trade-off relation in probabilistic entanglement generation

We investigate the generation of entanglement between two non-interacting systems by synthesizing a new quantum process from the superposition of distinct processes characterized by local-only operations. Our analysis leads to the derivation of a universal trade-off relation, $P_{\text{succ}}(1+\mathcal{C})\le1$, that fundamentally bounds the success probability ($P_{\text{succ}}$) and the generated entanglement (concurrence $\mathcal{C}$). The derivation of this trade-off relation is inspired by indefinite causal order, but applies for a broader class of quantum processes. Next, we show that the mathematical structure of this bound predicts the existence of a "quasi-deterministic" mode of operation, a surprising phenomenon which we then confirm with concrete entanglement generation protocols, where a maximally entangled state is guaranteed to be produced. In this mode of operation, both outcomes of the post-selection measurement on the auxiliary control system result in a maximally entangled state of the target system. Furthermore, we demonstrate how this general principle can be realized using a quantum switch, which leverages an indefinite causal order as a physical resource, and explore the rich variety of dynamical behaviors governed by the universal trade-off. Our results establish a general principle for entanglement generation with superposition of quantum processes and introduce a novel way of controlling entanglement generation.

quant-ph↗

arXiv2021

A simplified electro-chemical lithium-ion battery model applicable for in situ monitoring and online control

The penetrations of lithium-ion batteries in transport, energy and communication systems are increasing rapidly. A meticulous model applicable for precise in-situ monitoring and convenient online controlling is in sought to bridge the gap between research and applications. This paper proposes a simplified electro-chemical model and its discrete-time state-space realization derived from the pseudo-two-dimensional model. The solution-phase migration and solid-phase diffusion dynamics with varying parameters are captured and rigorous mathematical expressions of reaction rate distribution and terminal voltage are derived. A simulation framework including initializing, stabilizing and closed-loop correcting schemes with low computation cost are designed. Numeric experiments on different types of batteries in various operating scenarios are conducted for validation.

eess.SY↗

arXiv2021

Indefinite Causal Order in Quantum Batteries

Operations performing on quantum batteries are extended to scenarios where we no longer force the existence of definite causal order of occurrence between distinct processes. In contrast to standard theories, the so called indefinite causal order is found to have the capability of accomplishing tasks that are not possible without it. Specifically, we show how this novel class of resource comes into play in quantum batteries by first, combining two static unitary chargers into a coherently superposed one to fully charge an empty battery even if in the presence of battery's local Hamiltonian. Then we demonstrate for a non-unitary charging protocol, the indefinite causal order version charger yields a charged battery with higher energy over its classical counterpart under any conditions. We also have a finding that runs counter to our intuition which, roughly speaking, has the implication that a relatively less powerful charger guarantees a charged battery with higher energy. Finally, to reduce the cost imposed by a measurement-based protection scheme, indefinite causal order shows its potential to fulfill this goal.

quant-ph↗

arXiv2021

A novel energy resolved neutron imaging detector based on TPX3Cam for the CSNS

The China Spallation Neutron Source (CSNS) operates in pulsed mode and has a high neutron flux. This provides opportunities for energy resolved neutron imaging by using the TOF (Time Of Flight) approach. An Energy resolved neutron imaging instrument (ERNI) is being built at the CSNS but significant challenges for the detector persist because it simultaneously requires a spatial resolution of less than 100 {\mu}m, as well as a microsecond-scale timing resolution. This study constructs a prototype of an energy resolved neutron imaging detector based on the fast optical camera, TPX3Cam coupled with an image intensifier. To evaluate its performance, a series of proof of principle experiments were performed in the BL20 at the CSNS to measure the spatial resolution and the neutron wavelength spectrum, and perform neutron imaging with sliced wavelengths and Bragg edge imaging of the steel sample. A spatial resolution of 57 {\mu}m was obtained for neutron imaging by using the centroiding algorithm, the timing resolution was on the microsecond scale and the measured wavelength spectrum was identical to that measured by a beam monitor. In addition, any wavelengths can be selected for the neutron imaging of the given object, and the detector can be used for Bragg edge imaging. The results show that our detector has good performances and can satisfy the requirements of ERNI for detectors at the CSNS

physics.ins-det↗

arXiv2021

Feasibility study of TPC detector at high luminosity $Z$ pole on the circular collider

With the development of the circular collider, it is necessary to make accurate physics experimental measurements of particle properties at higher luminosity $Z$ pole. Micro-pattern gaseous detectors (MPGDs), which contain Gaseous Electron Multiplier (GEM) and Micro-mesh gaseous structures (Micromegas), have excellent potential for development as the readout devices of the time projection chamber (TPC) tracker detector. To meet the updated physics requirements of the high luminosity $Z$ from the preliminary concept design report (preCDR) to concept design report (CDR) at the circular electron positron collider (CEPC), In this paper, the space charge distortion of the TPC detector is simulated with the CEPC beam structure. Using the multi-physics simulation software package, the distribution of ion estimated by Geant4 is used as the input for the differential equation, and the relationship between the ion density distribution and electric field in the detector chamber is simulated. These simulation results show that the maximum deviation for Higgs $\mathcal{O}$(25 $\mu$m) meets the performance requirements in CEPC TPC detector at the high luminosity $Z$ pole, while it is still a considerable challenge for $Z$ pole, with the maximum deviation $\mathcal{O}$($>100~\mu$m). According to the previous developments, the cascaded structure of GEM and Micromegas detector has been measured. The new considerations of the detector's requirements were given, the gain needs to be reached to about 2000 with IBF$\times$Gain under 0.1, and IBF means the ions back flow ratio of the detector. The pixel TPC is a potential option to replace the traditional MPGDs with the low gain, low occupancy, and outstanding pattern recognition. Finally, some update parameters and experiments results were compared.

physics.ins-det↗

arXiv2020

A sealed ceramic GEM-based neutron detector

The GEM-based neutron detector has flourished in the past decade. However almost all the GEM-based neutron detectors work in the flow-gas mode, and the long-term performances of the detectors may be unstable due to the dynamic changes of atmospheric pressure and ambient temperature. In this paper, a sealed ceramic GEM-based neutron detector was developed at China Spallation Neutron Source (CSNS) and its sensitive area was 100 mm * 100 mm. The characterizations of the detector were presented and discussed, which included the plateau curve, the neutron beam profile, the neutron wavelength spectrum, the spatial resolution (FWHM: 2.77 mm), the two-dimensional (2D) imaging ability, the neutron detection efficiency and the counting rate instability (Relative Standard Deviation (RSD): 0.7%). The results show that the detector has good performances in sealed mode, and it can be used for the measurement of the direct neutron beam at CSNS.

physics.ins-det↗

arXiv2019

Feasibility study of TPC tracker detector for the circular collider

The discovery of a SM Higgs boson at the LHC brought about great opportunity to investigate the feasibility of a Circular Electron Positron Collider (CEPC) operating at center-of-mass energy of $\sim 240$ GeV, as a Higgs factory, with designed luminosity of about $2\times 10^{34}cm^{-2}s^{-1}$. The CEPC provides a much cleaner collision environment than the LHC, it is ideally suited for studying the properties of Higgs boson with greater precision. Another advantage of the CEPC over the LHC is that the Higgs boson can be detected through the recoil mass method by only reconstructing Z boson decay without examining the Higgs decays. In Concept Design Report(CDR), the circumference of CEPC is 100km, with two interaction points available for exploring different detector design scenarios and technologies. The baseline design of CEPC detector is an ILD-like concept, with a superconducting solenoid of 3.0 Tesla surrounding the inner silicon detector, TPC tracker detector and the calorimetry system. Time Projection Chambers (TPCs) have been extensively studied and used in many fields, especially in particle physics experiments, including STAR and ALICE. The TPC detector will operate in continuous mode on the circular machine. To fulfill the physics goals of the future circular collider and meet Higgs/$Z$ run, a TPC with excellent performance is required. We have proposed and investigated the ions controlling performance of a novel configuration detector module. The aim of this study is to suppress ion backflow ($IBF$) continually. In this paper, some update results of the feasibility and limitation on TPC detector technology R$\&$D will be given using the hybrid gaseous detector module.

physics.ins-det↗