SSL4Q: Semi-Supervised Learning of Quantum Data with Application to Quantum State Classification
Authors: Yehui Tang, Nianzu Yang, Mabiao Long, Junchi Yan
ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our empirical studies encompass simulations on two types of quantum systems: the Heisenberg Model and the Variational Quantum Circuit (VQC) Model, with system size reaching up to 50 qubits. The numerical results demonstrate SSL4Q s superiority over traditional supervised models in scenarios with limited labels, highlighting its potential in efficiently classifying quantum states with reduced computational and resource overhead. |
| Researcher Affiliation | Academia | 1School of Artificial Intelligence & Department of Computer Science and Engineering & Mo E Lab of AI, Shanghai Jiao Tong University, Shanghai, China. Correspondence to: Junchi Yan <yanjunchi@sjtu.edu.cn>. |
| Pseudocode | No | The paper describes the model architecture and training process using text and mathematical equations, but it does not contain any structured pseudocode or algorithm blocks that are explicitly labeled as 'Pseudocode' or 'Algorithm'. |
| Open Source Code | No | The paper mentions using third-party tools like 'Pennylane toolkit' and 'ITensor Julia implementation' but does not include any unambiguous statement of releasing its own source code for the described methodology, nor does it provide a direct link to such code. |
| Open Datasets | No | The paper describes how the quantum datasets for the Heisenberg and VQC models were generated (e.g., 'We generate a quantum dataset including images of class 0 and class 1...'). While the VQC dataset is based on MNIST, the generated quantum datasets themselves are not stated to be publicly available, and no links, DOIs, or repositories for these specific generated datasets are provided. |
| Dataset Splits | No | The paper explicitly defines training and testing datasets with their respective sizes (e.g., 'For the training dataset... we set N tr = 2000' and 'For the testing dataset... resulting in N te = 3000'). However, it does not mention a separate validation dataset split. |
| Hardware Specification | No | The paper discusses the computational complexity of simulating quantum systems and mentions using software toolkits like Pennylane and ITensor, but it does not provide any specific details about the hardware (e.g., GPU models, CPU types, or cloud computing instances) used for running the experiments. |
| Software Dependencies | No | The paper mentions using the 'Julia implementation provided by ITensor' and the 'Pennylane toolkit' for simulations. However, it does not specify the version numbers for these software components or any other key libraries or dependencies, which is crucial for reproducibility. |
| Experiment Setup | Yes | The key to this process are two hyper-parameters: the standard deviation (std) of Gaussian noise and the consistency weight λ. Based on empirical evidence, we set the Gaussian noise mean to 0 and std to 0.25, which consistently yields optimal predictive accuracy across various scenarios. The training duration is established at 500 epochs. We employ a tailored schedule for the consistency weight λ, learning rate, and the Exponential Moving Average (EMA) decay coefficient α, as depicted in the Fig. 9. Initially, λ is increased slowly... The learning rate ascends to a maximum of 0.001 and subsequently follows a cosine-like decline. The α coefficient adjusts in three phases... The Adam optimizer is used for optimization. For SSL4Q, the architecture comprises 4 heads, 2 layers, and a hidden dimension of 128. |