RegData

A collection of regression datasets.

Install

pip install regdata

Quick example

import regdata as rd
rd.set_backend('torch') # numpy, tf (numpy is default)
X, y, X_test = rd.Step().get_data() # Loads step function dataset

Features

• Simple API for quick benchmarking on various datasets.
• Get data in any framework: torch, tensorflow or numpy by setting a global backend.
• Scale X and/or y data with MinMaxScaler or StandardScaler.
• Get y in squeezed (n,) or unsqueezed (n,1) format.
• Perform only mean normalization on y.
• Add custom noise to the observations (y).
• Get consistent data with fixed random seed.

Plot datasets to have a quick glance

import regdata as rd
rd.Olympic().plot()

Checkout all plots here.

Datasets

from regdata import (
    DellaGattaGene,
    Heinonen4,
    Jump1D,
    MotorcycleHelmet,
    NonStat2D,
    Olympic,
    SineJump1D,
    SineNoisy,
    Smooth1D,
    Step
)

References

• DellaGattaGene
• Heinonen4
• Jump1D
• MotorcycleHelmet
• NonStat2D
• Olympic
• SineJump1D
• SineNoisy
• Smooth1D - Example 2
• Step