END-TO-END DEEP LEARNING CONSTRUCTION

Given the characteristics of white wine, the software predicts its quality in range of 10.

Table of Contents

• Run a hyperparameter sweep on a training script

• Compare the results of the runs in the MLflow UI

• Choose the best run and register it as a model (best-wine-quality model)

• Deploy the model to a REST API (future work)

• Build a container image suitable for deployment to a cloud platform (future work)

1. Hyperparameter Sweep

The investigated hyperparameters are learning rate and momentum. In the notebook, the selection in hyperparameter values are uniform.

The given loss function is Root Mean Square Error.

2. Result Comparison

The result from MLlfow UI is shown below.

3. Model Selection

The best model is one whose lr is 0.0074 and momentum is 0.92.