Quantum Implicit Neural Representations

Authors: Jiaming Zhao, Wenbo Qiao, Peng Zhang, Hui Gao

ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Lastly, we conducted experiments in signal representation, image superresolution, and image generation tasks to show the superior performance of QIREN compared to state-of-the-art (SOTA) models.
Researcher Affiliation Academia 1School of New Media and Communication, Tianjin University, China 2College of Intelligence and Computing, Tianjin University, China.
Pseudocode No The paper includes architectural diagrams (e.g., Figure 3) and mathematical expressions but does not contain any structured pseudocode or algorithm blocks.
Open Source Code Yes Our code is available at https://github.com/GGor MM1/QIREN.
Open Datasets Yes In the image representation task, we utilize three popular images: Astronaut, Camera, and Coffee (Van der Walt et al., 2014). [...] FFHQ is a high-resolution dataset of 70k human faces (Karras et al., 2019) and Celeb AHQ is a high-quality version of Celeb A that consists of 30k images (Karras et al., 2017).
Dataset Splits No The paper describes the datasets used and training details (e.g., number of epochs) but does not explicitly specify the division of data into training, validation, and test sets with percentages, sample counts, or specific citations to predefined splits for reproduction. For instance, it mentions downsampling images for experimentation but not the precise splitting strategy.
Hardware Specification No The paper states, "All our experiments are conducted on a simulation platform, utilizing the Pennylane (Bergholm et al., 2018) and Torch Qauntum (Wang et al., 2022)." It does not provide any specific details about the underlying hardware (e.g., CPU, GPU models, memory) used for this simulation platform.
Software Dependencies No The paper mentions using "Pennylane (Bergholm et al., 2018) and Torch Qauntum (Wang et al., 2022)". However, it does not provide specific version numbers for these software components, which is required for a reproducible description of software dependencies.
Experiment Setup Yes We use MSE as the loss function and use Adam optimizers with the parameters β1 = 0.9, β2 = 0.999 and ϵ = 1e 8. The models are trained for 600 epochs for Astronaut, 300 epochs for Camera, Coffee or sound. [...] QIREN includes three Hybrid layers and one output linear layer. Each Hybrid Layer consists of a linear layer with a hidden dimension of 8, a Batch Norm layer and a quantum circuit with 8 qubits.