Experimental Data for Nonlinear Time Series Forecasting via Recurrent Neural Network Optimized by Ising Machine

This dataset contains the experimental results and plotting data for the research paper. The provided data supports the comparative analysis of various recurrent neural network (RNN) training architectures. The dataset (data_plot.zip) includes: * Logistic Map Time-Series Data: The synthetic chaotic sequence generated with a bifurcation parameter r=3.65. It consists of a training set of 950 points and a testing set of 200 points. Model Prediction Results: Comparative experimental trajectories for the following methods: * Standard RNN based on Backpropagation Through Time (BPTT-RNN). * Reservoir Computing based on Ridge Regression (Ridge-RC). * The proposed Ising-based RNN (Ising-RNN) using Simulated Annealing (SA) and Simulated Ising Machine (SIM) solvers. * RMSE Statistical Data: The Root Mean Square Error (RMSE) values calculated as a function of prediction time steps for all evaluated models. Purpose: This data is released to ensure the reproducibility of the findings and to provide a benchmark for next-generation neural computing systems powered by Ising machines.本数据集收录了对应研究论文的实验结果与绘图数据，可为各类循环神经网络（Recurrent Neural Network, RNN）训练架构的对比分析提供支撑。 本数据集（data_plot.zip）包含以下内容： • 逻辑映射时序数据：采用分岔参数r=3.65生成的合成混沌序列，包含950个样本的训练集与200个样本的测试集。 • 模型预测结果：针对以下方法的对比实验轨迹： - 基于时间反向传播（Backpropagation Through Time, BPTT）的标准循环神经网络（BPTT-RNN） - 基于岭回归（Ridge Regression）的储备池计算（Reservoir Computing, RC） - 本文提出的采用模拟退火（Simulated Annealing, SA）与模拟伊辛机（Simulated Ising Machine, SIM）求解器的伊辛型循环神经网络（Ising-RNN） • 均方根误差统计数据：针对所有评估模型，计算得到的均方根误差（Root Mean Square Error, RMSE）值随预测时间步长的变化关系。 数据集发布目的：保障研究结论的可复现性，并为基于伊辛机的下一代神经计算系统提供性能基准。