Derivative of the MIMIC IV Waveform Database formatted to be suitable for machine learning. Formatting All records are split into intervals of roughly 60 seconds. The parameter values are averaged over each 60 second interval. The PPG signal data are unprocessed, i.e. as in the original dataset. Intervals with PPG signals containing missing data or large constant data are excluded. PPG signals and signal times are truncated to have the same amount of data points for all records. Formatted data are split into 3 different file types, namely *_n.csv containing the averaged parameter values, *_s.npy containing PPG signal data and *_t.npy containing the respective signal measurement times. Moreover, formatted data are split into trainXX_*, validation_* and test_* data files, where the training data trainXX_* are split into multiple files for easier handling. This dataset was created using the following code: https://gitlab.com/qumphy/wp1-benchmark-data-conversion Funding The creation of this dataset has been supported by the European Partnership on Metrology programme 22HLT01 QUMPHY. This project (22HTL01 QUMPHY) has received funding from the EMPIR programme cofinanced by the Participating States and from the European Union’s Horizon 2020 research and innovation programme.