Preprocessed ECG and Medical Image Datasets for Feedback-Free Learning

125.2 MB of NPZ and ZIP files contain preprocessed biomedical data from the PTB-XL ECG dataset and OCT/CXR medical image collections. These datasets support the experiments in the paper "Closed-Form Feedback-Free Learning with Forward Projection" by Robert O'Shea, updated in March 2026. The specific row and column counts are not detailed in the provided metadata.

Use Cases

• Train feedback-free learning models using preprocessed ECG signals from the PTB-XL dataset.
• Apply forward projection methods to preprocessed OCT and CXR image data for medical diagnosis tasks.
• Reproduce experiments from the associated paper using the provided NPZ and ZIP file formats.

Strengths

• Data is sourced from authoritative, publicly available biomedical datasets like PTB-XL.
• Includes 125.2 MB of preprocessed data in ready-to-use NPZ and ZIP formats.
• Associated code for full experiment reproduction is provided on GitHub.

Limitations

• Unknown sample size and feature dimensions limit assessment of dataset scale and suitability for specific models.
• Data is a preprocessed derivative; original data characteristics and any preprocessing transformations are not detailed.
• Temporal and geographic coverage of the source data is not specified.

Provenance

Source

Derived from PTB-XL ECG dataset and OCT/CXR collections by Kermany et al.

Collection Method

Preprocessed for experiments in "Closed-Form Feedback-Free Learning with Forward Projection".

Time Range

null

Freshness

Last updated March 14, III.

Geography

null