An interactive grammar of machine learning.
Project description
A high-level english representation of machine learning; modify what you want and let us handle the rest.
Installation
Install latest release version:
pip install -U nylon-ai
Install directory from github:
git clone https://github.com/Palashio/nylon.git
cd nylon-ai
pip install .
Usage: the basics
Nylon works through the nylonProcessor
object. Generally, a new object is creating everytime you're working with a new dataset. When initializing an object, a dataset in the form of a .csv or .xs file should be passed to it by path:
nylon_object = nylonProcessor('housing.csv')
Now, it's time to create a specifications file using the nylon grammar. Here's a basic one, that lets nylon handle most of the work. Nylon currently has four major parts in it's grammar: the data reader, preprocessor, modeler, and analysis modules. In the example below, you can see that we're specifying the target column under data (which is always required), and manually specifying the type of preprocessing we'd like. Everything we haven't specified will be handled for us.
{
"data": {
"target": "ocean_proximity"
},
"preprocessor": {
"fill": "ALL",
"label-encode": "ocean_proximity"
}
}
Now, we can override more components to take advantage of the built in ensembling of SVM's, and nearest neighbors modeling in nylon.
json_file = {
"data": {
"target": "ocean_proximity"
},
"preprocessor": {
"fill": "ALL",
"label-encode": "ocean_proximity"
},
"modeling": {
"type": ["svms", "neighbors"]
}
}
Now we can call,
nylon_object.run(json_file)
This will return a fully trained nylon object. You can access all information about this particular iteration in the .results
field of the object.
Asking for help
Welcome to the Nylon community!
If you have any questions, feel free to:
Contact
Shoot me an email at hello@paraglide.ai if you'd like to get in touch!
Follow me on twitter for updates and my insights about modern AI!
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.