Goal

This is the sharing session for my team, the goal is to quick ramp up the essential knowledges for linear regression case to experience how machine learning works during 1 hour. This sharing will recap basic important concepts, introduce runtime environments, and go through the codes on Notebooks of Azure Machine Learning Studio platform.

Recap of basic concepts

Do not worry about these theories if you can’t catch up, just take it as an intro.

Steps of machine learning

0. Get familiar with dataset, do preprocessing works.
1. Define the model, like linear model or neural network.
2. Define the goodness/cost of model, metrics can be error, cross entropy, etc.
3. Calculate the best function by optimization algorithms.

Linear model

Let’s start with the simplest linear model , you can also try more complex model if you get trouble in underfitting.

Question: How to initialize parameters?

Generalization

The model’s ability to adapt properly to new, previously unseen data, drawn from the same distribution as the one used to create the model.

• Underfitting: model is too simple to learn the underlying structure of the data (large bias)
• Overfitting: model is too complex relative to the amount and noisiness of the training data (large variance)...