pete88b / deep_learning_with_python

Study notes: Deep Learning with Python, Second Edition François Chollet

Study notes: Deep Learning with Python

Working through Second Edition François Chollet

Exploring fit_residual

The fit_residual idea is to build a model that has direct access to baseline results - in the hope that the model can learn to correct the errors of the baseline with a result that is;

• more accurate than without fit_residual and/or
• achieved with less compute / smaller models.

i.e. this is feature engineering and problem framing engineering to inject useful assumptions into the model.

• 10.2_temperature_forecasting.ipynb
  • creates a fully connect model that can do better than the non-machine learning baseline
• 10.2_temperature_forecasting_part2.ipynb
  • shows how fit_residual can be used to get a single layer LSTM to the same test MAE as stacked GRU
  • we also use the date features to improve MAE
• 10.2_california_housing.ipynb
  • shows that the fit_residual won't work on all datasets (o:

Anaconda environment creation

I use this for trying things out on my CPU-only windows machine.

conda create -n tf python==3.9 -y
conda activate tf
pip install tensorflow
pip install nbdev jupyterlab pandas matplotlib
pip install --upgrade jupyter notebook