Skip to content

Latest commit

 

History

History
82 lines (64 loc) · 2.84 KB

CHANGELOG.md

File metadata and controls

82 lines (64 loc) · 2.84 KB

hypers changelog

Version 0.0.12 (pre-release)

Features

  • Dataset object has been removed in favour of subclassing the numpy array object numpy.ndarray and extending its functionality for hyperspectral data. This forms the main data type for hypers and is hypers.hparray.

Version 0.0.11 (pre-release)

Features

  • Can now multiply, divide, add and subtract image array from data (e.g. (300,300,1000) * (300,300)) directly on Dataset
  • Printing the Dataset instance now prints information about the stored data

Version 0.0.10 (pre-release)

Features

  • Added MeanShift, AffinityPropagation and DBSCAN to the cluster method in Dataset
  • Added Normalizer to preprocess method in Dataset
  • Added following magic methods to Dataset:
    • setitem
    • mul
    • truediv
    • add
    • sub
  • Added plotting for cluster, decompose and scree
  • Added vertex component analysis as vca method in Dataset
  • Added unconstrained least-squares spectra fitting in abundance method in Dataset
  • Added Gaussian mixture models in the mixture method in Dataset

Performance enhancements

  • Looping over dims using ndindex, ndenumerate (speed improvement)

Version 0.0.9 (pre-release)

Features

  • Renamed package to hypers

Version 0.0.8 (pre-release)

Features

  • Added preprocess method to Process class which takes preprocessing classes from sklearn to preprocess data.

Version 0.0.7 (pre-release)

Features

  • Removed skhyper wrappers for scikit-learn classes. Added new decompose and cluster methods to the Process class to which a scikit-learn class is passed.
  • Removed skhyper wrappers for classification. This will be added again in a similar vein to the above decompose and cluster methods in the future.

Version 0.0.6 (pre-release)

Features

  • Added a plotting package for cluster/decomposition techniques and Process object

Performance enhancements

  • Moved normalization and smoothing directly to the Process class.

Bug fixes

  • Corrected issue with updating spectrum in the hyperspectral viewer for asymmetrical images.

Version 0.0.5 (pre-release)

Features

  • Added MLPClassifier in skhyper.neural_network

Miscellaneous

  • Added tests for MLPClassifier

Version 0.0.4 (pre-release)

Features

Performance enhancements

  • Decomposition techniques return a list of arrays for spec_components and image_components

Bug fixes

  • Corrected issue with output of predict() in reshaping array (issue with the classifiers)

Miscellaneous

  • Added tests for SVC, GaussianNB and KNeighborsClassifier
  • Changed test dataset from random to sklearn's make_blob

Version 0.0.3 (pre-release)

Features

  • Added GaussianNB in skhyper.naive_bayes
  • Added SVC in skhyper.svm
  • Added KNeighborsClassifier in skhyper.neighbors
  • Scree plot array can be accessed directly from Process object