Oracle Data Science Model Training

Model training builds the best mathematical representation of the relationship among features and a target label in the case of supervised learning and among all features in unsupervised learning.
It is an artifact created by the training process, which captures this relationship or pattern.
Determines the best algorithm for model training; considers tradeoffs in terms of computing, storage, complexity, performance, explainability, and so on

Training Process

This flow diagram represents the typical steps performed during model training.

There are a few functions at play here

Loss function: Also known as the cost function, compares the predictions made by the model (Ŷ) to the true target values(Y) in the training data set. The objective of the loss function is to represent the difference between the predicted values and the true target values as a single number, the loss score.

Optimizer: Uses the loss score to optimize the model weights. The goal of the optimizer is to improve the model’s accuracy without overfitting the training dataset.

Update: A function that updates the parameters of the fitted model from one iteration to the next.

The graph on the left represents this process as the green dots are the true data points. The black line through the middle represents predictions, and the red arrows represent the loss. An update function updates the parameters of the fitted model from one iteration to the next.

Model Training Libraries

Oracle Cloud Infrastructure Data Science notebooks can be used to train models by using either Oracle ADS or open source data science libraries, or a combination of both.

You can also install any external library by using a terminal in any existing environment if needed. Or you can just start with a set of your own choice of libraries and frameworks.