2 Getting started

 

This chapter covers

  • Installing Kubeflow and its components
  • Defining our first machine learning problems
  • Using Jupyter notebooks to explore data
  • Storing and accessing data from an object store.
  • Training a model and storing parameters for future use.
  • Deploying a model by creating a REST endpoint that can be queried for predictions.

In this chapter we will first explore the Kubeflow installation method and Kubeflow components to pave the way for installing Kubeflow and getting ready to use it. This chapter will also provide a simple example machine learning use case that is easy to follow and provides a high level exposure to different Kubeflow components such as Jupyter Notebooks, Minio object storage and Seldon for serving. The goal of this chapter is to provide readers with a “hello world” type of machine learning example that can be easily implemented in Kubeflow and can give users a quick entry point into Kubeflow.

2.1 Kubeflow installation introduction and its components

2.1.1 Kfctl and Kustomize

2.1.2 KfDef and Manifest

2.1.3 Choosing your Kubeflow Installation

2.2 Our First Example - Supervised Learning on a Synthetic Dataset

2.2.1 Our Problem Definition

2.3 Data Exploration and Development Environment - Jupyter Notebooks

2.3.1 Accessing notebooks used in this book

2.4 Data Storage and Ingestion

2.5 Model Training

2.5.1 Neural Networks: Some fundamentals

2.6 Model Deployment

2.7 Summary