Skip to main content

A probabilistic logic programming framework to represent and solve MDPs.

Project description

MDP-ProbLog is a Python3 framework to represent and solve (infinite-horizon) MDPs using Probabilistic Logic Programming.

Install

It is required to have Python3 installed.

$ pip3 install mdpproblog

Usage

$ mdp-problog --help
usage: mdp-problog {list, show, simulate, solve} [-m DOMAIN INSTANCE] [OPTIONS]

MDP-ProbLog is a Python3 framework to represent and solve Markovian Decision
Processes by Probabilistic Logic Programming. This project is free software.
Please check the documentation at http://pythonhosted.org/mdpproblog/.

positional arguments:
  {list,show,solve,simulate}
                        available commands: list examples, show and solve
                        models or simulate optimal policy

optional arguments:
  -h, --help            show this help message and exit
  -m MODEL MODEL, --model MODEL MODEL
                        list of domain and instance files
  -x EXAMPLE, --example EXAMPLE
                        select model from examples
  -g GAMMA, --gamma GAMMA
                        discount factor (default=0.9)
  -e EPSILON, --epsilon EPSILON
                        maximum error (default=0.1)
  -t TRIALS, --trials TRIALS
                        number of trials (default=100)
  -z HORIZON, --horizon HORIZON
                        simulation horizon (default=30)

Input

Domain specification for the sysadmin planning problem (models/sysadmin/domain.pl).

% Network topology properties
accTotal([],A,A).
accTotal([_|T],A,X) :- B is A+1, accTotal(T,B,X).
total(L,T) :- accTotal(L,0,T).
total_connected(C,T) :- connected(C,L),
                        total(L,T).

accAlive([],A,A).
accAlive([H|T],A,X) :- running(H,0), B is A+1, accAlive(T,B,X).
accAlive([H|T],A,X) :- not(running(H,0)), B is A, accAlive(T,B,X).
alive(L,A) :- accAlive(L,0,A).
total_running(C,R) :- connected(C,L),
                      alive(L,R).

% State fluents
state_fluent(running(C)) :- computer(C).

% Actions
action(reboot(C)) :- computer(C).
action(reboot(none)).

% Transition model
1.00::running(C,1) :- reboot(C).
0.05::running(C,1) :- not(reboot(C)), not(running(C,0)).
P::running(C,1)    :- not(reboot(C)), running(C,0),
                      total_connected(C,T), total_running(C,R), P is 0.45+0.50*R/T.

% Utility attributes

% costs
utility(reboot(C), -0.75) :- computer(C).
utility(reboot(none), 0.00).

% rewards
utility(running(C,0),  1.00) :- computer(C).

Example

$ mdp-problog simulate -x sysadmin1

Value(running(c1,0)=0, running(c2,0)=0, running(c3,0)=0) = 16.829
Value(running(c1,0)=1, running(c2,0)=0, running(c3,0)=0) = 19.171
Value(running(c1,0)=0, running(c2,0)=1, running(c3,0)=0) = 19.205
Value(running(c1,0)=1, running(c2,0)=1, running(c3,0)=0) = 23.028
Value(running(c1,0)=0, running(c2,0)=0, running(c3,0)=1) = 19.206
Value(running(c1,0)=1, running(c2,0)=0, running(c3,0)=1) = 23.029
Value(running(c1,0)=0, running(c2,0)=1, running(c3,0)=1) = 21.392
Value(running(c1,0)=1, running(c2,0)=1, running(c3,0)=1) = 25.607

Policy(running(c1,0)=0, running(c2,0)=0, running(c3,0)=0) = reboot(c1)
Policy(running(c1,0)=1, running(c2,0)=0, running(c3,0)=0) = reboot(c3)
Policy(running(c1,0)=0, running(c2,0)=1, running(c3,0)=0) = reboot(c1)
Policy(running(c1,0)=1, running(c2,0)=1, running(c3,0)=0) = reboot(c3)
Policy(running(c1,0)=0, running(c2,0)=0, running(c3,0)=1) = reboot(c1)
Policy(running(c1,0)=1, running(c2,0)=0, running(c3,0)=1) = reboot(c2)
Policy(running(c1,0)=0, running(c2,0)=1, running(c3,0)=1) = reboot(c1)
Policy(running(c1,0)=1, running(c2,0)=1, running(c3,0)=1) = reboot(none)

>> Value iteration converged in 0.196sec after 40 iterations.
>> Average time per iteration = 0.005sec.

Expectation(running(c1,0)=0, running(c2,0)=0, running(c3,0)=0) = 16.733
Expectation(running(c1,0)=1, running(c2,0)=0, running(c3,0)=0) = 19.433
Expectation(running(c1,0)=0, running(c2,0)=1, running(c3,0)=0) = 19.108
Expectation(running(c1,0)=1, running(c2,0)=1, running(c3,0)=0) = 23.377
Expectation(running(c1,0)=0, running(c2,0)=0, running(c3,0)=1) = 19.546
Expectation(running(c1,0)=1, running(c2,0)=0, running(c3,0)=1) = 23.287
Expectation(running(c1,0)=0, running(c2,0)=1, running(c3,0)=1) = 21.785
Expectation(running(c1,0)=1, running(c2,0)=1, running(c3,0)=1) = 25.849

License

Copyright (c) 2016-2017 Thiago Pereira Bueno All Rights Reserved.

MDPProbLog is free software: you can redistribute it and/or modify it under the terms of the GNU Lesser General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

MDPProbLog is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more details.

You should have received a copy of the GNU Lesser General Public License along with MDPProbLog. If not, see http://www.gnu.org/licenses/.

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

mdpproblog-0.3.0.tar.gz (28.1 kB view hashes)

Uploaded Source

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