sequence-scatter-link-neighborhood

• Temporal Sequence A deep learning training process consists of multiple training epochs. The predictions of the same data sample may vary across different epochs. Therefore, the predictions of all data in a specific epoch can be viewed as a "frame" in an animation. To display different frames in the same chart, we utilize the player component in vchart to automatically update epoch data, showcasing the progression of the training process over time.
• Scatter Plot To visualize the prediction results (in coordinate form) of each training data sample in the current epoch, it's natural to use a scatter plot where each data sample corresponds to a point in the chart. In addition to 2D coordinates, points can have other attributes, such as:
• Color: Can correspond to sample labels, with different labels having different colors;
• Opacity: Can correspond to the prediction confidence of samples, with higher confidence resulting in lower opacity;
• Size: Can distinguish selected points, with hovered or selected points enlarged;
• ......

For more implementation details, see: https://uvwoh700cpq.feishu.cn/docx/AULUdfgPJopMoMxP64RcuGo7nzh?from=from_copylink