Prompt Description

A schematic illustration depicts the training process for time series imputation. A grid-like matrix represents a multivariate time series, with rows indicating time steps and columns representing variables. The matrix contains missing values (represented as empty or light gray cells), and additional randomly masked cells (darker gray or transparent) to simulate missing data for training. The illustration distinguishes between observed values, original missing values, and artificially masked values used as prediction targets. Arrows or visual cues indicate that the model uses the partially observed matrix as input to reconstruct the masked values. The style is minimalistic, flat design, suitable for a scientific paper figure, with soft colors and high clarity, appropriate for a machine learning conference presentation. The illustration contains no text, labels, numbers, or annotations.