arXiv subjects

He Ma

Publications and source records attributed to He Ma.

At least 19 recordsLinked to original sources

arXiv2026

Dynamics Creation through Neural Dynamical Transfer Learning

Data-driven machine learning has established a robust foundation for reconstructing nonlinear dynamical systems from observations, primarily for the purposes of forecasting and control. However, most existing efforts focus on recovering specific observed dynamics rather than the generative synthesis of new ones. Inspired by image fusion and style transfer, we introduce a neural network framework termed Neural Dynamical Transfer Learning (NDTL) to create new systems with prescribed dynamics from pairs of parent nonlinear dynamical systems. By computing fundamental dynamical signatures, including the intrinsic dimension, the Kaplan-Yorke dimension, the invariant measure statistics, and the Lyapunov spectrum, we demonstrate that NDTL preserves key features inherited from the parent models while simultaneously generating novel dynamics. Beyond these validation examples, NDTL induces a criterion for dynamics classification, creates stable oscillatory coexistence in the Hastings-Powell food chain model, produces interpretable epidemiological models, and provides a chaotic source for image encryption.

nlin.CD↗

arXiv2026

TRWH: A Text-Driven Random Walk Heterogeneous GNN for Semantic-Aware Sparse Recommendation

Graph Neural Networks (GNNs) and Large Language Models (LLMs) have each advanced recommendation systems by modeling structural and semantic signals, respectively. However, integrating their complementary strengths remains challenging, particularly in sparse settings where maintaining semantic precision is critical. We propose TRWH (Text-driven Random Walk Heterogeneous Graph Neural Network), a novel framework that fuses LLM-generated textual profiles with heterogeneous graph structures through strategic random walk augmentation. TRWH consists of three core components: (1) Embedding Creation, which produces user and item representations using both Word2Vec and LLM-based profiling; (2) a Heterogeneous Graph Neural Network (HeteroGNN) that propagates information across multi-relational edges; and (3) Random Walk-based Path Construction, which enriches sparse graphs with second-order user-user and item-item links. Experiments on the Amazon-2023 Fashion (2M users, 825K items) and Beauty (631K users, 112K items) datasets demonstrate that TRWH achieves substantial performance gains over state-of-the-art methods, including 80.0% RMSE and 52.6% MAE reductions on Fashion, and 25.7% and 10.8% improvements on Beauty. Notably, while random walks improve performance with traditional embeddings, they can dilute the nuanced representations learned by LLMs, underscoring the importance of adaptive integration strategies.

cs.AI↗

arXiv2026

RecEvolve: A Knowledge-Driven Autonomous Agent System for Recommender Systems

The rise of agentic AI has catalyzed a shift toward self-iterating systems, opening new frontiers for the autonomous optimization of production recommender models. This paper presents the empirical validation of a knowledge-driven autonomous agent system, deployed directly on a production large-scale Two-Tower retrieval model. By delegating the entire research lifecycle, spanning idea generation, code implementation, offline training, and metric evaluation, to a continuous closed-loop autonomous framework, the agent system executed over 40 completed autonomous training runs from scratch. Executing these runs under rigorous production-scale evaluations, the system systematically navigated hidden architectural bottlenecks on the latest production model to achieve a breakthrough ~20% relative improvement in NDCG, a gain that translated directly to a +3.77% increase in user satisfaction in live production traffic. Furthermore, the deployment exposed critical vulnerabilities in standard evaluation protocols, as the agent system autonomously discovered reward-hacking shortcuts. These findings prove that an autonomous pipeline can dramatically accelerate the pace of machine learning research and stress-test the rigorousness of underlying experimental infrastructure, while also exposing novel challenges such as reward hacking and redundant exploration of failed hypotheses.

cs.IR↗

arXiv2025

PLUM: Adapting Pre-trained Language Models for Industrial-scale Generative Recommendations

Large Language Models (LLMs) pose a new paradigm of modeling and computation for information tasks. Recommendation systems are a critical application domain poised to benefit significantly from the sequence modeling capabilities and world knowledge inherent in these large models. In this paper, we introduce PLUM, a framework designed to adapt pre-trained LLMs for industry-scale recommendation tasks. PLUM consists of item tokenization using Semantic IDs, continued pre-training (CPT) on domain-specific data, and task-specific fine-tuning for recommendation objectives. For fine-tuning, we focus particularly on generative retrieval, where the model is directly trained to generate Semantic IDs of recommended items based on user context. We conduct comprehensive experiments on large-scale internal video recommendation datasets. Our results demonstrate that PLUM achieves substantial improvements for retrieval compared to a heavily-optimized production model built with large embedding tables. We also present a scaling study for the model's retrieval performance, our learnings about CPT, a few enhancements to Semantic IDs, along with an overview of the training and inference methods that enable launching this framework to billions of users in YouTube.

cs.IR↗

arXiv2025

User Feedback Alignment for LLM-powered Exploration in Large-scale Recommendation Systems

Exploration, the act of broadening user experiences beyond their established preferences, is challenging in large-scale recommendation systems due to feedback loops and limited signals on user exploration patterns. Large Language Models (LLMs) offer potential solutions by leveraging their world knowledge to recommend novel content outside these loops. A key challenge is aligning LLMs with user preferences while preserving their knowledge and reasoning. To enhance planning for new user interests using LLMs, this paper introduces a novel approach that combines hierarchical planning with LLM inference-time scaling. This method aims to improve recommendation relevancy without compromising novelty. We decouple novelty and user-alignment, training separate LLMs for each objective. We then scale up the novelty-focused LLM's inference and select the best-of-n predictions using the user-aligned LLM. Live experiments demonstrate efficacy, showing significant gains in both user satisfaction (measured by watch activity and active user counts) and exploration diversity.

cs.IR↗

arXiv2024

LLMs for User Interest Exploration in Large-scale Recommendation Systems

Traditional recommendation systems are subject to a strong feedback loop by learning from and reinforcing past user-item interactions, which in turn limits the discovery of novel user interests. To address this, we introduce a hybrid hierarchical framework combining Large Language Models (LLMs) and classic recommendation models for user interest exploration. The framework controls the interfacing between the LLMs and the classic recommendation models through "interest clusters", the granularity of which can be explicitly determined by algorithm designers. It recommends the next novel interests by first representing "interest clusters" using language, and employs a fine-tuned LLM to generate novel interest descriptions that are strictly within these predefined clusters. At the low level, it grounds these generated interests to an item-level policy by restricting classic recommendation models, in this case a transformer-based sequence recommender to return items that fall within the novel clusters generated at the high level. We showcase the efficacy of this approach on an industrial-scale commercial platform serving billions of users. Live experiments show a significant increase in both exploration of novel interests and overall user enjoyment of the platform.

cs.IR↗

arXiv2024

Weakly supervised alignment and registration of MR-CT for cervical cancer radiotherapy

Cervical cancer is one of the leading causes of death in women, and brachytherapy is currently the primary treatment method. However, it is important to precisely define the extent of paracervical tissue invasion to improve cancer diagnosis and treatment options. The fusion of the information characteristics of both computed tomography (CT) and magnetic resonance imaging(MRI) modalities may be useful in achieving a precise outline of the extent of paracervical tissue invasion. Registration is the initial step in information fusion. However, when aligning multimodal images with varying depths, manual alignment is prone to large errors and is time-consuming. Furthermore, the variations in the size of the Region of Interest (ROI) and the shape of multimodal images pose a significant challenge for achieving accurate registration.In this paper, we propose a preliminary spatial alignment algorithm and a weakly supervised multimodal registration network. The spatial position alignment algorithm efficiently utilizes the limited annotation information in the two modal images provided by the doctor to automatically align multimodal images with varying depths. By utilizing aligned multimodal images for weakly supervised registration and incorporating pyramidal features and cost volume to estimate the optical flow, the results indicate that the proposed method outperforms traditional volume rendering alignment methods and registration networks in various evaluation metrics. This demonstrates the effectiveness of our model in multimodal image registration.

cs.CV↗

arXiv2023

Imaging the Meissner effect and flux trapping in a hydride superconductor at megabar pressures using a nanoscale quantum sensor

By directly altering microscopic interactions, pressure provides a powerful tuning knob for the exploration of condensed phases and geophysical phenomena. The megabar regime represents an exciting frontier, where recent discoveries include novel high-temperature superconductors, as well as structural and valence phase transitions. However, at such high pressures, many conventional measurement techniques fail. Here, we demonstrate the ability to perform local magnetometry inside of a diamond anvil cell with sub-micron spatial resolution at megabar pressures. Our approach utilizes a shallow layer of Nitrogen-Vacancy (NV) color centers implanted directly within the anvil; crucially, we choose a crystal cut compatible with the intrinsic symmetries of the NV center to enable functionality at megabar pressures. We apply our technique to characterize a recently discovered hydride superconductor, CeH$_9$. By performing simultaneous magnetometry and electrical transport measurements, we observe the dual signatures of superconductivity: local diamagnetism characteristic of the Meissner effect and a sharp drop of the resistance to near zero. By locally mapping the Meissner effect and flux trapping, we directly image the geometry of superconducting regions, revealing significant inhomogeneities at the micron scale. Our work brings quantum sensing to the megabar frontier and enables the closed loop optimization of superhydride materials synthesis.

cond-mat.supr-con↗

arXiv2022

Evolving symbolic density functionals

Systematic development of accurate density functionals has been a decades-long challenge for scientists. Despite the emerging application of machine learning (ML) in approximating functionals, the resulting ML functionals usually contain more than tens of thousands parameters, which makes a huge gap in the formulation with the conventional human-designed symbolic functionals. We propose a new framework, Symbolic Functional Evolutionary Search (SyFES), that automatically constructs accurate functionals in the symbolic form, which is more explainable to humans, cheaper to evaluate, and easier to integrate to existing density functional theory codes than other ML functionals. We first show that without prior knowledge, SyFES reconstructed a known functional from scratch. We then demonstrate that evolving from an existing functional $\omega$B97M-V, SyFES found a new functional, GAS22 (Google Accelerated Science 22), that performs better for the majority of molecular types in the test set of Main Group Chemistry Database (MGCDB84). Our framework opens a new direction in leveraging computing power for the systematic development of symbolic density functionals.

cs.NE↗

arXiv2022

Noisy Sensor Scheduling in Wireless Networked Control Systems: Freshness or Precision

In linear wireless networked control systems whose control is based on the system state's noisy and delayed observations, an accurate functional relationship is derived between the estimation error and the observations' freshness and precision. The proposed functional relationship is then applied to formulate and solve the problem of scheduling among different wireless links from multiple noisy sensors, where a sliding window algorithm is further proposed. The algorithm's simulation results show significant performance gain over existing policies, even in scenarios that require high freshness or precision of observations.

cs.IT↗

arXiv2021

First-principles predictions of out-of-plane group IV and V dimers as high-symmetry high-spin defects in hexagonal boron nitride

Hexagonal boron nitride (h-BN) has been recently found to host a variety of quantum point defects, which are promising candidates as single-photon sources for solid-state quantum nanophotonics applications. Most recently, optically addressable spin qubits in h-BN have been the focus of intensive research due to their unique potential in quantum computing, communication, and sensing. However, the number of high-symmetry high-spin defects that are desirable for developing spin qubits in h-BN is highly limited. Here, we combine density functional theory (DFT) and quantum embedding theories to show that out-of-plane XY dimer defects (X, Y = C, N, P, Si) form a new class of stable C3v spin-triplet defects in h-BN. We find that the dimer defects have a robust 3A2 ground state and 3E excited state, both of which are isolated from the h-BN bulk states. We show that 1E and 1A shelving states exist and they are positioned between the 3E and 3A2 states for all the dimer defects considered in this study. To support future experimental identification of the XY dimer defects, we provide an extensive characterization of the defects in terms of their spin and optical properties. We predict that the zero-phonon line of the spin-triplet XY defects lies in the visible range (800 nm - 500 nm). We compute the zero-field splitting of the dimers to range from 1.79 GHz (SiP) to 29.5 GHz (CN). Our results broaden the scope of high-spin defect candidates that would be useful for the development of spin-based solid-state quantum technologies in two-dimensional hexagonal boron nitride.

cond-mat.mtrl-sci↗

arXiv2021

Quantum Embedding Theory for Strongly-correlated States in Materials

Quantum embedding theories are promising approaches to investigate strongly-correlated electronic states of active regions of large-scale molecular or condensed systems. Notable examples are spin defects in semiconductors and insulators. We present a detailed derivation of a quantum embedding theory recently introduced, which is based on the definition of effective Hamiltonians. The effect of the environment on a chosen active space is accounted for through screened Coulomb interactions evaluated using density functional theory. Importantly, the random phase approximation is not required and the evaluation of virtual electronic orbitals is circumvented with algorithms previously developed in the context of calculations based on many-body perturbation theory. In addition, we generalize the quantum embedding theory to active spaces composed of orbitals that are not eigenstates of Kohn-Sham Hamiltonians. Finally, we report results for spin defects in semiconductors.

cond-mat.mtrl-sci↗

arXiv2021

Spin-spin interactions in solids from mixed all-electron and pseudopotential calculations $-$ a path to screening materials for spin qubits

Understanding the quantum dynamics of spin defects and their coherence properties requires accurate modeling of spin-spin interaction in solids and molecules, for example by using spin Hamiltonians with parameters obtained from first-principles calculations. We present a real-space approach based on density functional theory for the calculation of spin-Hamiltonian parameters, where only selected atoms are treated at the all-electron level, while the rest of the system is described with the pseudopotential approximation. Our approach permits calculations for systems containing more than 1000 atoms, as demonstrated for defects in diamond and silicon carbide. We show that only a small number of atoms surrounding the defect needs to be treated at the all-electron level, in order to obtain an overall all-electron accuracy for hyperfine and zero-field splitting tensors. We also present results for coherence times, computed with the cluster correlation expansion method, highlighting the importance of accurate spin-Hamiltonian parameters for quantitative predictions of spin dynamics.

cond-mat.mtrl-sci↗

arXiv2020

Temperature evolution of quasiparticle dispersion dynamics in semimetallic 1T-TiTe2 via high-resolution angle-resolved photoemission spectroscopy and ultrafast optical pump-probe spectroscopy

High-resolution angle-resolved photoemission spectroscopy and ultrafast optical pump-probe spectroscopy were used to study semimetallic 1T - TiTe2 quasiparticle dispersion and dynamics. A kink and a flat band, having the same energy scale and temperature-dependent behaviors along the G-M direction, were detected. Both manifested at low temperatures but blurred as temperature increased. The kink was formed by an electron-phonon coupling. And the localized flat band might be closely related to an electron-phonon coupling. Ultrafast optical spectroscopy identified multiple distinct time scales in the 10-300 K range. Quantitative analysis of the fastest decay process evidenced a significant lifetime temperature dependence at high temperatures, while this starts to change slowly below ~ 100 K where an anomalous Hall coefficient occurred. At low temperature, a coherent A1g phonon mode with a frequency of ~ 4.36 THz was extracted. Frequency temperature dependence suggests that phonon hardening occurs as temperature falls and anharmonic effects can explain it. Frequency fluence dependence indicates that the phonons soften as fluence increases.

cond-mat.mtrl-sci↗

arXiv2020

Probing the coherence of solid-state qubits at avoided crossings

Optically addressable paramagnetic defects in wide-band-gap semiconductors are promising platforms for quantum communications and sensing. The presence of avoided crossings between the electronic levels of these defects can substantially alter their quantum dynamics and be both detrimental and beneficial for quantum information applications. Avoided crossings give rise to clock transitions, which can significantly improve protection from magnetic noise and favorably increase coherence time. However, the reduced coupling between electronic and nuclear spins at an avoided crossing may be detrimental to applications where nuclear spins act as quantum memories. Here we present a combined theoretical and experimental study of the quantum dynamics of paramagnetic defects interacting with a nuclear spin bath at avoided crossings. We develop a computational approach based on a generalization of the cluster expansion technique, which can account for processes beyond pure dephasing and describe the dynamics of any solid-state spin-qubits near avoided crossings. Using this approach and experimental validation, we determine the change in nature and source of noise at avoided crossings for divacancies in SiC. We find that we can condition the clock transition of the divacancies in SiC on multiple adjacent nuclear spins states. In our experiments, we demonstrate that one can suppress the effects of fluctuating charge impurities with depletion techniques, leading to an increased coherence time at clock transition, limited purely by magnetic noise. Combined with ab-initio predictions of spin Hamiltonian parameters, the proposed theoretical approach paves the way to designing the coherence properties of spin qubits from first principles.

quant-ph↗

arXiv2020

First-principles Studies of Strongly Correlated States in Defect Spin Qubits in Diamond

Using a recently developed quantum embedding theory, we present first principles calculations of strongly correlated states of spin defects in diamond. Within this theory, effective Hamiltonians are constructed, which can be solved by classical and quantum computers; the latter promise a much more favorable scaling as a function of system size than the former. In particular, we report a study of the neutral group-IV vacancy complexes in diamond, and we discuss their strongly-correlated spin-singlet and spin-triplet excited states. Our results provide valuable predictions for experiments aimed at optical manipulation of these defects for quantum information technology applications.

cond-mat.mtrl-sci↗

arXiv2020

PyCDFT: A Python package for constrained density functional theory

We present PyCDFT, a Python package to compute diabatic states using constrained density functional theory (CDFT). PyCDFT provides an object-oriented, customizable implementation of CDFT, and allows for both single-point self-consistent-field calculations and geometry optimizations. PyCDFT is designed to interface with existing density functional theory (DFT) codes to perform CDFT calculations where constraint potentials are added to the Kohn-Sham Hamiltonian. Here we demonstrate the use of PyCDFT by performing calculations with a massively parallel first-principles molecular dynamics code, Qbox, and we benchmark its accuracy by computing the electronic coupling between diabatic states for a set of organic molecules. We show that PyCDFT yields results in agreement with existing implementations and is a robust and flexible package for performing CDFT calculations. The program is available at https://github.com/hema-ted/pycdft/.

cond-mat.mtrl-sci↗

arXiv2020

Entanglement and control of single quantum memories in isotopically engineered silicon carbide

Nuclear spins in the solid state are both a cause of decoherence and a valuable resource for spin qubits. In this work, we demonstrate control of isolated 29Si nuclear spins in silicon carbide (SiC) to create an entangled state between an optically active divacancy spin and a strongly coupled nuclear register. We then show how isotopic engineering of SiC unlocks control of single weakly coupled nuclear spins and present an ab initio method to predict the optimal isotopic fraction which maximizes the number of usable nuclear memories. We bolster these results by reporting high-fidelity electron spin control (F=99.984(1)%), alongside extended coherence times (T2=2.3 ms, T2DD>14.5 ms), and a >40 fold increase in dephasing time (T2*) from isotopic purification. Overall, this work underlines the importance of controlling the nuclear environment in solid-state systems and provides milestone demonstrations that link single photon emitters with nuclear memories in an industrially scalable material.

quant-ph↗