As a non-programmer, I have a question about ml_regression_train.
When using ml_deformer, it is the result of a fully connected training. This may lead to an issue where the range of skeletal animation influence is too large.such as hand bone movements potentially affecting the skin dynamics of the legs. Therefore, a weight parameter or penalty function is needed. Which would be more reasonable to use: adding weights or applying a penalty?

Additionally, since this involves hundreds of bone parameters, it would result in a very large array. How should this be stored more efficiently?

Another issue is that the training examples have undergone PCA processing. How should this be correlated with the bone weight parameters?