publications

2025

1. QCE25

   QSeer: A Quantum-Inspired Graph Neural Network for Parameter Initialization in Quantum Approximate Optimization Algorithm Circuits

   Lei Jiang, Chi Zhang, and Fan Chen

   2025

   arXiv
2. QCE25

   Qracle: A Graph-Neural-Network-based Parameter Initializer for Variational Quantum Eigensolvers

   Chi Zhang, Lei Jiang, and Fan Chen

   2025

   arXiv Code

2024

1. QCE24

   

   Entangled Meanings: Classification and Ambiguity Resolution in Near-Term QNLP

   Chi Zhang, Akriti Kumari, and Damir Cavar

   In 2024 IEEE International Conference on Quantum Computing and Engineering (QCE), Sep 2024

   Abs DOI Poster

   We demonstrate two tasks in Quantum Natural Language Processing (QNLP): classification and disambiguation. For the classification task, we utilized an amplitude encoding algorithm and achieved perfect accuracy on the standard lambeq dataset, which is commonly used for benchmarking in quantum NLP. Additionally, we obtained accuracy from 55% to 72.5 % on a more complex and realistic Amazon review dataset, which is still a reasonable result, given the current capability of quantum computing and QNLP. Additionally, we found that UMAP is the best dimension reduction method in our experiment setting. All classification results were done on the default. qubit simulator in pennylane 0.36 python library. Our classification results highlight the potential of quantum algorithms in practical applications. For the disambiguation task, we selected 18 ambiguous nouns, 32 unambiguous nouns, and 18 different verbs. Our experiments using the QASM simulator within the qiskit Python library demonstrated that the simulator could perfectly differentiate between the various meanings of ambiguous nouns in different contexts, achieving perfect differentiation. Furthermore, we extended our study to a real quantum device, the ibm_kyoto quantum computer. There, we tested our disambiguation approach on a subset of 4 random nouns (two ambiguous and two unambiguous) and observed that ibm_kyoto could achieve an accuracy range of 82.1 % to 98.9% in disambiguation tasks, extending the datasets and improving the results of existing ambiguity resolution experiments [1]. Our work demonstrates the capability of quantum computing in dealing with real-world NLP tasks, hence contributing to the advancement of both QML and NLP.

2. QCE24

   

   Semantic Similarities Using Classical Embeddings in Quantum NLP

   Damir Cavar and Chi Zhang

   In 2024 IEEE International Conference on Quantum Computing and Engineering (QCE), Sep 2024

   Abs DOI Poster

   We demonstrate how word and text embeddings as n-dimensional real vectors used in classical Natural Language Processing (NLP)/AI applications can be mapped to quantum states or quantum embedding representations in Quantum Applications using NLP and AI. Models like fastText [1], [2], GloVe [3], Numberbatch [4], or BERT [5] are common NLP embedding vectors or models that encode semantic properties of words or text fragments. We use these models to evaluate mapping properties and compression rate, as well as information preservation in quantum embeddings. For mapping embedding vectors to quantum states, we use encoding strategies, such as Amplitude Encoding. The encoding strategies allow us to map large dense vectors from the embedding models to compact quantum states using different compression ratios. The compression with Amplitude Encoding can be 2^n to N, resulting in a mapping of a 1,024-dimensional vector of reals in the classical environment to a 10-qubit state. The goal of this work is to evaluate these strategies with respect to their compression ratio and measure the preservation of semantic information using similarity scores. We show that the resulting quantum embeddings based on mapped classical computing embeddings exhibit the same relational properties and that there is no significant loss of semantic information in the conversion from classical n-dimensional real vector embeddings to qubit states. We conclude that the experimental results allow us to quantum compute semantic similarities of words or text, reusing existing and freely available embedding models from classical NLP/AI computing.

3. QCE24

   Entangled Meanings: Classification and Ambiguity Resolution in QNLP

   Chi Zhang, Akriti Kumari, and Damir Cavar

   In 2024 IEEE International Conference on Quantum Computing and Engineering (QCE), Sep 2024

   Abs DOI Code

   We discuss experiments involving two tasks in Quantum Natural Language Processing (QNLP): text classification and disambiguation. In the classification task, we utilized an amplitude encoding algorithm and achieved perfect accuracy on the lambeq dataset discussed in literature. We obtained accuracy from 55% to 72.5% on the more complex and realistic Amazon review dataset. This is a reasonable result given the current state-of-the-art results in QNLP. Additionally, when using vector dimension reduction for embeddings, we found that UMAP leads to the best results in our experiment setting. All classification results were done on the default. qubit simulator in pennylane 0. 36 python library. Our classification results highlight the potential of quantum algorithms in practical applications. In the disambiguation task, we selected 18 ambiguous nouns, 32 unambiguous nouns, and 18 different verbs. Our experiments using the QASM simulator within the qiskit Python library demonstrated that the simulator could perfectly differentiate between the various meanings of ambiguous nouns in different contexts. Furthermore, we extended our study to a real quantum device, the ibm_kyoto quantum computer. There, we tested our disambiguation approach on a subset of 4 random nouns (2 ambiguous and 2 unambiguous) and observed that ibm_kyoto could achieve an accuracy range of 82.1% to 98.9% in disambiguation tasks, extending the datasets and improving the results of existing ambiguity resolution experiments [1]. Our work demonstrates the capability of quantum computing in dealing with real-world NLP tasks, hence contributing to the advancement of both OML and NLP.

2022

1. Comms. Theor. Phys.

   Four-family N =1 supersymmetric Pati–Salam models from intersecting D6-branes

   Tianjun Li, Rui Sun, and Chi Zhang

   Communications in Theoretical Physics, Jun 2022

   Abs DOI arXiv

   We investigate the construction of four-family supersymmetric Pati–Salam models from Type IIA orientifold with intersecting D6-branes. Utilizing the deterministic algorithm introduced in He W, Li T and Sun R (2021 arXiv: hep-th/2112.09632), we obtain 274 types of models with three rectangular tori and distinct gauge coupling relations at string scale, while 6 types of models with two rectangular tori and one tilted torus. In both cases, there exists a class of models with gauge coupling unification at string scale. In particular, for the models with two rectangular tori, one tilted torus and gauge coupling unification, the gaugino condensations are allowed, and thus supersymmetry breaking and moduli stabilization are possible for further phenomenological study.