Skip to main content

Python library for manipulating probabilistic automata.

Project description

Probabilistic Automata

Build Status Docs codecov PyPI version License: MIT

Python library for manipulating Probabilistic Automata. This library builds upon the dfa package.


If you just need to use probabilistic_automata, you can just run:

$ pip install probabilistic_automata

For developers, note that this project uses the poetry python package/dependency management tool. Please familarize yourself with it and then run:

$ poetry install


The probabilistic_automata library centers around the PDFA object which models a finite probabilistic transition system, e.g., a Markov Decision Process, as a DFA or Moore Machine over a product alphabet over the system's actions and the environment's stochastic action.

import probabilistic_automata as PA

def transition(state, composite_action):
    sys_action, env_action = composite_action
    return (state + sys_action + env_action) % 2

def env_dist(state, sys_action):
    """Based on state and the system action, what are the probabilities 
    of the environment's action."""

    return {0: 1/2, 1: 1/2}  # Always a coin flip.

noisy_parity = PA.pdfa(
    inputs={0, 1},
    env_inputs={0, 1},
    outputs={0, 1},
    env_dist=env_dist,   # Equivalently, PA.uniform({0, 1}).

The support and transition probabilities can easily calculated:

assert, 0) == {0, 1}
assert noisy_parity.transition_probs(0, 0) == {0: 1/2, 1: 1/2}
assert noisy_parity.prob(start=0, action=0, end=0) == 1/2

Dict <-> PDFA

Note that pdfa provides helper functions for going from a dictionary based representation of a probabilistic transition system to a PDFA object and back.

import probabilistic_automata as PA

mapping = {
    "s1": (True, {
        'a': {'s1': 0.5, 's2': 0.5},
    "s2": (False, {
        'a': {'s1': 1},

start = "s1"
pdfa = PA.dict2pdfa(mapping=mapping, start=start)
assert pdfa.inputs == {'a'}

mapping2, start2 = PA.pdfa2dict(pdfa)
assert start == start2
assert mapping2 == mapping


The probabilistic_automata library has two convenience methods for transforming a Deterministic Finite Automaton (dfa.DFA) into a PDFA.

  • The lift function simply creates a PDFA whose transitions are deterministic and match the original dfa.DFA.
import probabilistic_automata as PA
from dfa import DFA

parity = DFA(
    inputs={0, 1},
    transition=lambda s, c: (s + c) & 1,

parity_pdfa = lift(parity)

assert pdfa.inputs == PARITY.inputs
assert pdfa.env_inputs == {None}
  • The randomize function takes a DFA and returns a PDFA modeling the actions of the DFA being selected uniformly at random.
pdfa = PA.randomize(PARITY)

assert pdfa.inputs == {None}
assert pdfa.env_inputs == PARITY.inputs

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

probabilistic_automata-0.1.1.tar.gz (7.5 kB view hashes)

Uploaded source

Built Distribution

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page