# Examples¶

Here are examples on how to use MDP for typical machine learning applications:

- Logistic Maps — Using Slow Feature Analysis (SFA) for processing a non-stationary time series, derived by a logistic map.
- Growing Neural Gas — Capture the topological structure of a data distribution.
- Locally Linear Embedding — Approximate data with a low-dimensional surface and reduce its dimensionality by learning a mapping to the surface.
- Fast image filtering using the caching extension — Filter images with 2D wavelets and demonstrate use of caching extension.
- Handwritten digits classification with MDP and scikits.learn — Use the combined power of MDP and scikits.learn in an applciation for handwritten digit classification
- hinet_html.py — Get the HTML representation for a simple hinet network.
- benchmark_parallel.py — Simple benchmark to compare the different schedulers in MDP.
- pp_remote_test.py — Simple test of the remote Parallel Python support, using the NetworkPPScheduler.
- Slideshow and Double slideshow — Created slideshows of matplotlib plots, demonstrates the slideshow module in MDP.
- hinetplaner — Interactive HTML/JS/AJAX based GUI for constructing special hinet networks. This is a complicated example which won’t teach you much about MDP.
- mnist — Several more example for handwritten digit classification, this time with Fisher Discriminant Analysis and without scikits.learn.

The following examples use and illustrate BiMDP.

- BiMDP flow inversion — A simple example on the alternative mechanism to inverse a BiFlow.
- BiMDP Hinet Inspection — Demonstrates the inspection of a BiFlow.
- BiMDP Custom Inspection — Customization with maptlotlib plots of the BiFlow inspection.
- bimdp_simple_coroutine.py — Minimal example for coroutine decorator.
- Gradient Descent and the Newton Method — Use Newton’s method for gradient descent with the gradient extension.
- Backpropagation — Implement backpropagation for a multi layer perceptron.
- Deep Belief Network (DBN) based on BiMDP — Proof of concept for a Deep Belief Network.