DeepLearningForTimeSeriesForecasting/hyperparameter_tuning

Hyper-parameter tuning using Azure Machine Learning service

The hyper-parameters of methods presented in this tutorial are tuned using Hyperdrive, a feature of Azure Machine Learning (Azure ML) service.

Pre-Requisites

• You'll need to first set up the environment using Anaconda3. To do so, navigate to the hyperparameter_tuning directory and run the following commands:

  # create a conda environment using the environment.yaml file
  # replace 'create' with 'update' if you need to update the environment
  conda env create -f environment.yaml
  source activate dnntutorial
  # to install the environment into the Jupyter kernel, run
  python -m ipykernel install --user --name dnntutorial
  

• Next, please follow instructions in configuration notebook in order to provision Azure ML workspace.

Note: This material is developed on an Azure Data Science Virtual Machine (Ubuntu). This machine is pre-installed with many development tools we use in this tutorial, such as Anaconda, Python, etc.

Run hyper-parameter tuning notebook

Hyper-parameter tuning is done in hyperparameter_tuning.ipynb notebook. This notebook is used to tune several approaches:

• Feed-forward network multi-step multivariate - ff_multistep_config.json
• RNN multi-step - rnn_multistep_config.json
• RNN teacher forcing - rnn_teacher_forcing_config.json
• RNN encoder decoder - rnn_encoder_decoder_config.json
• CNN - cnn_config.json

Each of these use cases is defined in a json configuration file listed above alongside each usecase. To run a specific approach, please specify the appropriate configuration file in the hyperparameter_tuning notebook.

The running time depends on the size of your Azure ML cluster and the model being tuned.