lucamarx / pyautospec

Simple spectral learning for weighted automata

pyAutoSpec

pyAutoSpec is a simple library for demonstrating the spectral learning algorithm detailed in the 2014 paper Spectral learning of weighted automata by Borja Balle, Xavier Carreras, Franco M. Luque and Ariadna Quattoni.

You can find an introduction on my blog.

Disclaimer

This is a toy implementation of spectral learning:

• it is not optimized
• it is not thoroughly tested
• it is as simple as possible

DO NOT USE IN PRODUCTION

If you need spectral learning in your application use scikit-splearn instead.

Installation

To build the library just run

pip install -e .

Or use the provided docker file:

docker build -t pyautospec .
docker run -p 8888:8888 -it pyautospec

and access the Jupyter notebook.

Quick Start

Suppose you have a function f that takes a string and returns a number, you want to learn a weighted finite automaton that computes f, you do this

from pyautospec import SpectralLearning

alphabet = [ord(c) for c in range(ord('a'), ord('z')+1)]

learner = SpectralLearning(alphabet, 3)

automaton = learner.learn(f)

see the Automata Learning example.

The same thing can be done for a function f that takes a number in an interval and returns a number, in this case you can do this

from math import sin, pi
from pyautospec import FunctionWfa

# learn the sin function in the [0,2π] interval
sin_aut = FunctionWfa().fit(sin, x0=0.0, x1=2*pi, learn_resolution=2)

see the Trigonometric Functions example.