Phase Field Raw Data

This dataset is the raw data for the dataset found at https://data.dtu.dk/articles/dataset/Phase_field_data/25562364
It comprises tailored phase field prediction data generated using an innovative automated workflow designed to offer insights into complex phenomena while minimizing computational expenses. The dataset aims to facilitate benchmarking of new algorithms in phase field prediction, emphasizing accessibility and utility for researchers. The data creation process is detailed, focusing on streamlining data collection and preparation. Validation of the dataset's effectiveness is conducted through a benchmark experiment utilizing U-Net regression, a widely adopted neural network architecture. Results showcase competitive performance of the U-Net model, akin to previous state-of-the-art methods. This dataset not only serves as a valuable resource for the phase field prediction community but also highlights the potential of U-Net regression, fostering further advancements in the field.
The linked code can be found under https://github.com/laura-rieger/phase_field_benchmark and describes in detail how the dataset is to be used.

本数据集为https://data.dtu.dk/articles/dataset/Phase_field_data/25562364 处公开数据集的原始数据。其包含定制化的相场（phase field）预测数据，该数据通过创新的自动化工作流生成，旨在在降低计算开销的同时，为复杂现象的研究提供科学洞见。本数据集旨在助力相场预测领域新算法的基准测试，同时强调对研究人员的易用性与实用价值。数据集的构建流程已详细说明，核心在于简化数据收集与预处理环节。本数据集的有效性通过基准实验得到验证：实验采用了当前广泛使用的神经网络架构U-Net回归。实验结果表明，U-Net模型取得了可与此前前沿最优方法媲美的性能。本数据集不仅为相场预测领域的研究者提供了宝贵的研究资源，同时也彰显了U-Net回归的应用潜力，有助于推动该领域的进一步发展。
相关代码可在https://github.com/laura-rieger/phase_field_benchmark 获取，其中详细说明了本数据集的使用方法。