Skip to content
This repository


Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP

Rapid Machine Learning Prototyping in Python

branch: master

This branch is 0 commits ahead and 0 commits behind master

Fetching latest commit…

Cannot retrieve the latest commit at this time

Ramp - Rapid Machine Learning Prototyping

Ramp is a python library for rapid prototyping of machine learning solutions. It's a light-weight pandas-based machine learning framework pluggable with existing python machine learning and statistics tools (scikit-learn, rpy2, etc.). Ramp provides a simple, declarative syntax for exploring features, algorithms and transformations quickly and efficiently.


Why Ramp?

  • Clean, declarative syntax

    No more hackish one-off spaghetti scripts!

  • Complex feature transformations

    Chain and combine features:

Interactions([Log('x1'), (F('x2') + F('x3')) / 2])
Reduce feature dimension:
DimensionReduction([F('x%d'%i) for i in range(100)], decomposer=PCA(n_components=3))
Incorporate residuals or predictions to blend with other models:
Residuals(config_model1) + Predictions(config_model2)
  • Data context awareness

    Any feature that uses the target ("y") variable will automatically respect the current training and test sets. Similarly, preparation data (a feature's mean and stdev, for example) is stored and tracked between data contexts.

  • Caching

    Ramp caches and stores on disk in fast HDF5 format (or elsewhere if you want) all features and models it computes, so nothing is recomputed unnecessarily. Results are stored and can be retrieved, compared, blended, and reused between runs.

  • Easy extensibility

    Ramp has a simple API, allowing you to plug in estimators from scikit-learn, rpy2 and elsewhere, or easily build your own feature transformations, metrics, feature selectors, reporters, or estimators.

Quick start

Getting started with Ramp: Classifying insults

Or, the quintessential Iris example:

import pandas
from ramp import *
import urllib2
import sklearn
from sklearn import decomposition

# fetch and clean iris data from UCI
data = pandas.read_csv(urllib2.urlopen(
data = data.drop([149]) # bad line
columns = ['sepal_length', 'sepal_width', 'petal_length', 'petal_width', 'class']
data.columns = columns

# all features
features = [FillMissing(f, 0) for f in columns[:-1]]

# features, log transformed features, and interaction terms
expanded_features = (
    features +
    [Log(F(f) + 1) for f in features] +
        F('sepal_width') ** 2,

# Define several models and feature sets to explore,
# run 5 fold cross-validation on each and print the results.
# We define 2 models and 4 feature sets, so this will be
# 4 * 2 = 8 models tested.


    # Try out two algorithms

    # and 4 feature sets

        # Feature selection
            # use random forest's importance to trim
            target=AsFactor('class'), # target to use
            n_keep=5, # keep top 5 features

        # Reduce feature dimension (pointless on this dataset)

        # Normalized features
        [Normalize(f) for f in expanded_features],


Ramp is very alpha currently, so expect bugs, bug fixes and API changes.


  • Numpy
  • Scipy
  • Pandas
  • PyTables
  • Sci-kit Learn


Ken Van Haren. Email with feedback/questions: @squaredloss

Something went wrong with that request. Please try again.