Hi,

I made a VAE encoder to do data augmentation of a tabular data.

I learned the model and I am able to augment my data giving these 4 input features :

x1 x2 x3 y

The new augmented data are pretty nice regarding the data it can generate and my real values.

Now, I would like to use this model to predict y values based on new incoming x1 x2 x3 values . I don’t really know which way to go ?

is there any relation between x1, x2, x3 and y tabular data?

Also can you briefly describe what are these 4 input features? do they share same dimensionality?

The data are correlated.

They can be seen as some parameters of metal deformations when a force is a applied

x1 may be a classifier as stainless steel, copper, aluminimum

x2 the metal sheet length

x3 a force applied to this metal sheet

(x4) y : a corresponding springback angle

if we suppose always the same metal sheet depth, for a given metal material

and a fixed metal length (x2), there exists a correlation between the force applied and it’s springback.

If I change the length of metal sheet (eg from 10cm to 1meter) , the springback for a same angle will not be the same , but keeps correlated to the length.

Changing the material I will have similar behaviour but with different data scales

augmented generated data with small sets of real data my VAE seems consistent with reality.

Now, I would like to be able to predict y=f(x1,x2,x3) using real and augmented data.

Can you share screenshots of this result ??