Hello
I am searching for a way to have 2D coordinate X and Y vector converted into n-dimensional vector. The transformation needs to be reversible.
The transformation can be a matrix multiplication (1X2 * 2Xn-d = 1Xn-d) but in that case the inverse of the chosen matrix is something that is not easily found because the matrix is not a square matrix. In that case transformation is not reversible.
Is there a method to do this kind of transformation without creating a function that passes through all values ?!
Cheers
Hello Me in the past
You can use an Autoencoder to find the solution. If you use linear layers you will find the parameters of the linear layer of the encoder form a transformation matrix and the linear layer params of the decoder represent the inverse of that matrix. Their multiplication is equal to identity.
Cordially
- Badges
- Science topic
- Similar topics
- Mathematics
- Algebra
- Matrix