Skip to content

Conceptual issue in autoencoder inference step #1

@padriba

Description

@padriba

The current inference section appears to contain a conceptual error regarding the use of the autoencoder output.

Current code:

encoded_data = model.predict(x_test)
decoded_data = model.predict(encoded_data)

Issue:
model.predict(x_test) already returns the reconstructed images produced by the autoencoder. Therefore, the variable encoded_data does not contain encoded latent vectors; it actually contains decoded/reconstructed images.

As a result, the following line:

decoded_data = model.predict(encoded_data)

feeds reconstructed images back into the autoencoder a second time, which is conceptually incorrect and may lead to misleading results.

Correct approach:

decoded_data = model.predict(x_test)

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions