AIModels package

Submodules

• AIModels.AIClasses module
  • Classes
  • Field
  • EarlyStopping
  • TimeSeriesDataset
  • PositionalEmbedding
    • PositionalEmbedding.forward()
  • FixedPositionalEmbedding
    • FixedPositionalEmbedding.forward()
  • FeaturePositionalEmbedding
    • FeaturePositionalEmbedding.forward()
  • SymmetricFeatureScaler
    • SymmetricFeatureScaler.fit()
    • SymmetricFeatureScaler.transform()
    • SymmetricFeatureScaler.inverse_transform()
  • IdentityScaler
    • IdentityScaler.fit()
    • IdentityScaler.transform()
    • IdentityScaler.inverse_transform()
• AIModels.AIutil module
  • Utility Treatment Routines for ML Predicting models
    • Utilities
  • func_name()
  • copy_dict()
  • get_arealat_arealon()
  • select_field()
  • select_field_key()
  • select_field_eof()
  • select_area()
  • make_matrix()
  • make_eof()
  • make_field()
  • make_field_V5()
  • make_field_HAD()
  • normalize_training_data()
  • make_data()
  • make_features()
  • make_data_base()
  • init_weights()
  • count_parameters()
  • epoch_time()
  • create_time_features()
  • rescale()
  • matrix_rank_light()
  • CPRSS()
  • transform_strings()
  • make_fcst_array()
  • eof_to_grid()
  • advance_months()
  • project_dyn()
  • make_dyn_verification_new()
  • compute_increments()
  • cumsum_with_init()
  • select_fcst()
  • variance_features()
  • create_subdirectory()
  • eof_to_grid_new()
  • get_common_dates()
• AIModels.AutoEncoder module
  • Encoder
    • Encoder.forward()
  • Decoder
    • Decoder.forward()
  • AutoEncoder
    • AutoEncoder.reparameterize()
    • AutoEncoder.forward()
  • train_autoencoder()
• AIModels.ClimFormer module
  • ClimFormer class, a subclass of InformerForPrediction
  • TimeSeriesDataset
  • TimeSeriesFuture
  • ClimFormer
  • TrasFormer
• AIModels.ClimLSTM module
  • TimeSeriesDataset
  • ClimLSTM
    • ClimLSTM.forward()
    • ClimLSTM.train_model()
    • ClimLSTM.evaluate_model()
    • ClimLSTM.predict()
• AIModels.LocalInformer module
  • Modified version of the InformerForPrediction class
    • Main Modifications:
  • InformerModel
    • InformerModel.get_lagged_subsequences()
    • InformerModel.create_network_inputs()
    • InformerModel.get_encoder()
    • InformerModel.get_decoder()
    • InformerModel.forward()
  • InformerForPrediction
    • InformerForPrediction.output_params()
    • InformerForPrediction.get_encoder()
    • InformerForPrediction.get_decoder()
    • InformerForPrediction.output_distribution()
    • InformerForPrediction.forward()
    • InformerForPrediction.generate()
• AIModels.ModelTraining module
  • Training, validation, and prediction methods for the Informer model.
    • Functions:
  • train_model()
  • validate_model()
  • predict()
  • deter_generate()
  • greedy()
• AIModels.UtilPlot module
  • Auxiliary Plotting routines
  • Single_Forecast_plots()
  • many_plots()
  • Forecast_plots()
  • Three_Forecast_plots()
  • select_months()
  • plot_skill()
  • extract_and_merge_csv()
  • write_skill_to_csv()
  • plot_csv()
  • boxplot()
  • write_var_excel()
  • Two_Forecast_plots()
  • write_var_table()
  • define_defaults_values()
  • get_common_dates()
  • calculate_significance()

Module contents

AIModels is a Python package for the designing ML networks of atmospheric and ocean data. It is based on the xarray package.

It uses xarray as a basic data structure.

AIModels contains several modules with classes and utility routines The plotting modules are based on matplotlib and cartopy. The computation modules are based on numpy and scipy.

ClimFormer :

Class for ClimFormer network

ClimLSTM :

Class for ClimLSTM network

LocalInformer :

Class for LocalInformer network, a modified version of Informer for time series from HuggingFace

ModelTraining :

Class for training, validate and inference the models

UtilPlot :

Routines for plotting and visualization of the data

AIutil :

Utilities for the rest of the modules

Version 2.1