I hope that there are some methods to show the prediction results of certain dataset. Those datasets have ten different labels. So the prediction results of one model also have ten point. I found some methods to draw point with three dim. However, I did not know how to convert the 10-D point to 3-D point reasonably. My 3D Model
CodePudding user response：
You have two options:
- Plot only three of the ten dimensions
You can just choose the first three or random three dimensions. But you will lose information about the other seven dimensions.
- Apply dimensionality reduction
You can reduce dimensions by some ML-Algorithm like t-SNE. This way you can visualize higher dimensional data e.g. in 3D without losing significant information. t-SNE is implemented in sklearn.
Hope this helps.
CodePudding user response：
Like leon said, you can apply dimensionality reduction to visualize the results of your model. A cool tool to do this is in this github repo.
https://github.com/harveyslash/TSNE-UMAP-Embedding-Visualisation
You can extract an embedding vector from your model, the predictions and the classes and look at how the data is displayed in a 3 dimensional graph. This graph can also display the raw images with the numbers in it so you can see how the model is grouping together the data.
It's really cool.