louisdorard/BML_course.md

## BML_course.md

      
    Raw
  

              BML_course.md
            
          
    Bootstrapping Machine Learning course

Baseline: Learn to use Machine Learning quickly with prediction services and APIs
Duration: 3 days
Who should take this course?


Companies looking to exploit the value of their data and to use them for decision making
Developers/hackers who want to acquire new skills and to learn to start building predictive apps
CTOs/Lead Developers who wish to understand the possibilities offered by Machine Learning and what is at stake
Innovators/startuppers who want to pursue opportunities in the world of Big Data
Analysts who would like to use Prediction services to enrich their recommendations

Description/overview

This course will teach you the basics of Machine Learning. At the end of the 3 days you will be autonomous in setting up predictive models. You will be able to create these models from data and to use them in your domain, thanks to Prediction services such as BigML and Google Prediction API. All along, you will put theory into practice by creating yourself a predictive app.
“Bootstrapping Machine Learning” - Complete Package included


You will receive Bootstrapping Machine Learning in paperback and in PDF/ePub/Mobi. If you wish, you will be able to read the book prior to the course. At the end of the course you will also receive the contents of the Complete Package (worth $299) which contains screencasts, IPython notebooks, code, datasets, a Virtual Machine, videos and other ressources.
Prerequisites


Basic knowledge of programming
Familiarity with spreadsheet programs (e.g. Microsoft Excel)

What should you bring?


Your own laptop

What will you learn?


Possibilities and limitations of Machine Learning
Formulating your own Machine Learning problem
Exploring and wrangling data
Building and using predictive models through Prediction APIs
Evaluating models' performance and impact
Deploying your own Specialized Prediction API

Program

Day 1: Learning from data


Introduction to Machine Learning and its possibilities:

Basic concepts
Examples in mobile and web apps
Examples of Data Science in Business


Revisions: basics of the Python scripting language
Using the BigML prediction service:

Overview of the web dashboard
Creating and using predictive models
Accessing BigML programmatically through the API and its Python wrappers


Data browsing within a spreadsheet program (Google Spreadsheets)
Using Google Prediction within Spreadsheets
Overview of the limitations of Machine Learning
Introduction to feature engineering
How to process data before learning predictive models from it:

Data wrangling and cleaning with Pandas
Pre-processing textual data
Data anonymization


Day 2: Evaluation and comparaison of predictive models


Using Google Prediction API:

Uploading data to Google Cloud Storage
Using the API through the web-based Google Cloud Console
Authentication procedure
Accessing the API programmatically through its Python wrappers


Evaluation procedures prior to deploying predictive models:

Splitting data into training and test sets
Testing data representativeness
Measures of predictive models' performance: accuracy, precision, recall, MSE, R-squared
Cross-validation procedure for reliable evaluations


Deployment, integration of a model, and measure of its impact

Day 3: Going further with predictive models


Deployment of your own specialized prediction API, on the cloud or on a private infrastructure
Advanced features of BigML: ensembles, clustering, Virtual Private Cloud
How to use predictive models to build recommender systems
Overview of the other Prediction APIs: Wise.io, Wit.ai, etcML, Semantria, Alchemy, etc.
Recap
Ressources to go further and to adapt Machine Learning solutions to your own needs