Analyzing Convergence in Quantum Neural Networks: Deviations from Neural Tangent Kernels
Authors: Xuchen You, Shouvanik Chakrabarti, Boyang Chen, Xiaodi Wu
ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments: sublinear QNN convergence To support Theorem 3.2, we simulate the training of QNNs using M0 with eigenvalues {±1}. |
| Researcher Affiliation | Collaboration | 1Department of Computer Science, University of Maryland, College Park, United States 2Global Technology Applied Research, J. P. Morgan Chase & Co. 3Institute for Interdisciplinary Information Sciences, Tsinghua University, Beijing, China. |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. Methods are described in prose and mathematical formulations. |
| Open Source Code | No | The paper does not provide concrete access to source code (e.g., a specific repository link or explicit code release statement) for the methodology described. |
| Open Datasets | No | A m-sample dataset is generated by randomly sampled m orthogonal pure states {vi}m j=1 Cd and randomly assigned half of the samples with label +1 and the other half label -1 (i.e. {yi}m j=1 {±1}m). |
| Dataset Splits | No | The paper does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) needed to reproduce the data partitioning into train/validation/test sets. |
| Hardware Specification | Yes | We run the experiments on Amazon EC2 C5 Instances. The simulation of the asymptotic dynamics is run on Intel Core i7-7700HQ Processor (2.80Ghz) with 16G memory. |
| Software Dependencies | No | We simulate the QNN experiments using Pytorch (Paszke et al., 2019). The paper mentions PyTorch but does not provide a specific version number or other software dependencies with version numbers. |
| Experiment Setup | Yes | To simulate the dynamics of gradient flow, we choose the learning rate to be 0.001/p and the maximum number of epochs is set to be 10000. |