Learn Browserify, Scikit-learn and Xstream

Artificial intelligence. Machine learning. Bots. Computers learning and communicating with us to do our bidding. But, where do you start? How do you get a machine to even begin to understand what you speak or type at it? There are several common machine learning algorithms that will help us begin to answer these questions.

In this course we’ll learn about common machine learning algorithms that don’t require implementing a neural network. We will not be going too much into the math behind them, but instead learn what each algorithm is good for, and how to train them. We'll also learn about a few metrics for evaluating models.

We’ll implement these in Python using `scikit-learn` using `scikit-learn`’s built-in data sets. The focus of this course is on implementation and a high-level understanding of these algorithms. 

We'll look at a few ways to evaluate our models, for both classification and regression models. We'll touch on mean squared error and coefficient of determination (for regression), and accuracy score, logarithmic loss, confusion matrices, and classification reports (for classification). 

Python 2.7 is used in the lesson videos but the code provided has Python 3 available. The only breaking change is the `print` statement API.

For additional information on installation, vocabulary, and common errors visit the `README.md` to the course code attached to each lesson.

Simple and efficient tools for data mining and data analysis

scikit-learn

Introductory Machine Learning Algorithms in Python with scikit-learn

Cycle.js is a framework where your app is described as a simple function taking an event stream as input and outputting an event stream.

cycle

This lesson is the first in an introduction to Cycle.js. We will see how to build from scratch a toy version of Cycle.js. See how to use RxJS to create simple web apps, and how to separate logic from effects.

Separate Logic from Effects in Cycle.js

Browserify lets you require('modules') in the browser by bundling up all of your dependencies.

browserify

Building upon the `watch` task that we begin this lesson with, we’ll look at how to create a local development server using the `app` directory as the web root. By using Browsersync to achieve this, we can also instruct all connected browsers to automatically reload each time our bundle is re-generated. Browsersync has a public `.stream()` method that is designed exactly for this purpose.

Gulp and Browserify - Adding Live Reload with Browsersync

Gulp is a toolkit for automating painful or time-consuming tasks in your development workflow, so you can stop messing around and build something.

gulp

What is already a nice workflow using Gulp, Browserify, Watchify & Browsersync to enable auto rebuilds & live browser reloading can be further improved by adding Babelify. A simple wrapper around the popular transpiler Babel, Babelify can ‘transform’ source files as they pass through our stream to enable certain Javascript features not yet available in the Browser or Node. We end the lesson by looking at how to generate & extract sourcemaps using `exorcist`.

Gulp and Browserify - Adding Babel & Source Maps

Let's look at the basics of setting up Gulp and Browserify. Creating a Gulp task from scratch, we'll cover how to utilise the Browserify API to generate a single `bundle.js` file, catch and log any errors that occur in the compilation process, transform the stream into a format that can be consumed by `gulp.dest` and finish by writing the bundle to a `dist` directory.

Gulp and Browserify - Initial Setup

Beginning with a single Gulp task capable of generating a one-time Browserify bundle, we enable file-watching and incremental rebuilds by incorporating Watchify. We add a second task `watch` & refactor our initial code to enable the bulk of it to be re-used. We’ll look at exactly how to ‘wire’ these tools together and how the relationship between Watchify & Browserify can manifest into a seriously efficient workflow.

Gulp and Browserify - Hooking up Watchify

Linear regression is a linear model that is used for regression problems, or problems where the goal is to predict a value on a continuous spectrum (as opposed to a discrete category).

We’ll use linear regression to estimate continuous values. In this case, we’ll predict house prices in Boston. We'll also look at how to visualize our results with matplotlib, and how to evaluate our models with different metrics for regression problems.

Use Linear Regression To Estimate Continuous Values with Python and Scikit-learn

Despite its often confusing name, logistic regression is a **linear** model that is used for **classification**, or estimating discrete values.

We'll use an inbuilt scikit-learn dataset of iris data to classify irises into three categories. We'll also look at metrics and tools to evaluate our classification models, including the accuracy score, classification report, and confusion matrix.

Predict Categories in Python using Scikit-learn's Logistic Regression module

We’ll continue with the iris dataset to implement support vector machines, which can be used to find more complex boundaries for classification or regression problems. 

More about Support Vector Machines can be found at [scikit-learn](http://scikit-learn.org/stable/modules/svm.html).

There are several types of kernel function that can be used with SVMs. Scikit-learn supports these kernels:

-linear

-polynomial ('poly')

-rbf (radial basis function) 

-sigmoid 

Custom kernels are also supported. `Rbf` is the default kernel type.

A good overview of kernels can be found [here](https://www.kdnuggets.com/2016/06/select-support-vector-machine-kernels.html), or at the [scikit-learn page](http://scikit-learn.org/stable/modules/svm.html#svm-kernels).

Use Support Vector Machines To Find Complex Decision Boundaries with Scikit-learn

We’ll continue with the iris dataset to implement k-nearest neighbors (`KNN`), which makes predictions about data based on similarity to other data instances. We'll visualize how the `KNN` algorithm works by making its predictions based on its neighbors' labels. We'll also examine the confusion matrix a bit further.

`KNN` can be used for both classification and regression problems. 

`KNN` is good for low dimensional data (data without too many input variables). It is not good for unbalanced data sets, and it can be computationally expensive.

Use K-nearest Neighbors To Find Similar Datapoints with Python and Scikit-learn

We’ll learn about decision trees, also known as CART (classification and regression trees), and use them to explore a dataset of breast cancer tumors. 

We'll also see how to visualize a decision tree using `graphviz`.

Implement Decision Trees in Python with Scikit-learn

We’ll return to the iris dataset to see how to use k-means clustering, an unsupervised learning algorithm, to create  categories for data that doesn't have labels. We'll also visualize these clusters using matplotlib.

More on K-means can be found at [Scikit-learn](http://scikit-learn.org/stable/modules/clustering.html#k-means).

Topics

Instructors

Content Type