Demos & Tutorials

MLDB ships with some demo and tutorial Notebooks.

How-to Tutorials

First steps with MLDB

• Using pymldb Tutorial
• Loading Data Tutorial
• Querying Data Tutorial
• SELECT Tutorial
• Procedures and Functions Tutorial
• Using pymldb Progress Bar and Cancel Button Tutorial

Intermediate

• Loading Data From An HTTP Server Tutorial
• Virtual Manipulation of Datasets Tutorial
• Executing JavaScript Code Directly in SQL Queries Using the jseval Function Tutorial
• Selecting Columns Programmatically Using Column Expressions Tutorial
• Identifying Biased Features Tutorial
• Tensorflow Image Recognition Tutorial

Benchmarks

The Benchmarking MLDB notebook shows how to run the The Absolute Minimal Machine Learning Benchmark with MLDB.

Transfer Learning with Tensorflow

The Transfer Learning with Tensorflow demo demonstrates how to do transfer learning to leverage the power of a deep convolutional neural network without having to train one yourself. Most people do not train those types of networks from scratch because of the large data and computational power requirements. What is more common is to train the network on a large dataset (unrelated to our task) and then leverage the representation it learned.

Image Processing and Image Recognition Basics

This topic is broken down in two demos. First, the Image Processing with Convolutions demo explains convolutions and shows different ways to do them with MLDB. Second, the Real-Time Digits Recognizer demo goes though the machine learning concepts necessary to build the MLPaint plugin.

Predicting Titanic Survival

The Predicting Titanic Survival demo shows a classification workflow:

• importing data with the import.text procedure type
• to training and evaluating a classifier with the classifier.experiment procedure type
• calling a classifier with the classifier function type
• understanding it with the classifier.explain function type

Mapping Reddit and Visualizing StackOverflow tags

The Mapping Reddit and Visualizing StackOverflow Tags demos show how to use MLDB to visualize high-dimensional datasets:

• doing dimensionality reduction with the svd.train procedure type
• clustering with the kmeans.train procedure type
• visualizting with the tsne.train procedure type.

Recommending Movies

The Recommending Movies demo shows how to use MLDB to do recommendation with the svd.train procedure type.

Exploring Favourite Recipes

The Exploring Favourite Recipes demo shows how to use MLDB to do clustering/topic extraction among recipes with the kmeans.train procedure type.

Spam Filtering

The Enron Spam Filtering demo uses MLDB to show the perils of over-reliance on Area Under the Curve as a metric for evaluating classifiers.

Natural Language Processing with Word2Vec

The Mapping Election Press Releases demo uses MLDB to visualize the relationships between texts with Word2Vec using the import.word2vec procedure type.

Using SQL to explore the Panama Papers

The Investigating the Panama Papers demo shows off MLDB's SQL engine by exploring the raw data from the Offshore Leaks Database.