AI library

F

Felix Benner

I thought about an AI library for python. This is the possible
structure I came up with. Are there any thoughts about it?

ailib/

search.py

class State:
"represents an immutable state of a problem"
def __str__(self):
pass

def __hash__(self):
pass

class StateSpace:
"represents a traversable collection of states"
def start(self):
"returns a list of initial states"
pass

def follow(self, state).
"returns a list of states that follow after state"
pass

def search(self, order=none):
"""returns an iterator over all states according to
order
order can be a constant (depth, breadth, incdepth) or
a
heuristic. If none some default search will be
used."""
pass

def __iter__(self):
return self.search()

class Heuristic:
def follow(self, state):
"returns a list of states probably closer to the
solution than s
pass

class Graph(StateSpace):
def __init__(self, paths)
"paths is a set of tuples (nodeFrom, nodeTo, weight)"
pass

def a-star(self, nodeFrom, nodeTo):
"""searches the shortest path (minimal weight) from
nodeFrom to nodeTo."""
pass

plan.py

class State(search.State):
"represents a state of the world that can be changed
through action.
pass

class Agent(search.StateSpace):
"""For any given state an Agent has a set of possible
actions
that transform the initial state into a subsequent
state."""

def perceive(self):
"returns the state of the world the agent perceives to
be in."
pass

def actions(self, state):
"returns an iterator over actions available in the
given state."
pass

def __iter__(self):
return self.actions(self.perceive())

def plan(self, state, currentState=none):
"""returns a sequence of actions that are supposed to
transform the currently perceived state of the world
into the desired state."""
if currentState==none:
currentState = self.perceive()
else:
pass

logic.py

class Symbol:
"a Symbol that can be bound to a value or unbound."
def bound(self):
"returns true if the symbol is bound."

def value(self):
"if bound returns the value."

class Function:
"transforms a list of symbols into another symbol."
def __init__(self, *symbols):
self.symbols = symbols

def __call__(self):
"returns some symbol."
pass

class Predicate:
"either is or is not valid for a given list of symbols"
def __init__(self, *symbols):
self.symbols = symbols

def __call__(self):
"returns true or false"
pass

class Junctor:
"""a relation between predicates deriving a truth value
from the
truth values of the predicates"""
def __init__(self, *predicates):
self.predicates = predicates

def __call__(self):
"returns some truth value"
pass

class Quantifier:
"somehow binds symbols."
pass

class Clause:
"A quantified junctor."
pass

class Axioms:
"A list of clauses"
def consistent(self):
"returns true if the list of axioms is consistent"
pass

def valid(self, clause, bind=true):
"""returns true if the clause is consistent with the
set of axio
If bind is true, any unbound symbol will be bound to a
value if
pass

statistics.py

class Entity:
"bearer of statistically relevant features."
pass

class Set:
"set of entities. defines usual statistical functions"
def __iter__(self):
"iterate over all entities."

def avg(self, feature):
"returns the average of the given feature of all
entities."
pass
# likewise other functions

def entropy(self, a, b):
"returns the level of randomnes (0..1) between the
given feature
pass

def correlates(self, a, b, e=0.5):
return self.entropy(a, b) < e

def mine(self):
"returns the tuple of features with the least
entropy."

game.py

class Game:
"""Basically extends planning with multiple agents who
take turns."""
pass

language.py

class Morphology:
"word recognition."
def lookup(self, word, guess=false):
"""returns a list of dictionaries of features for the
given word
If guess is true, tries to guess features if
unsure."""

class Rule:
"a Rule of the language like S -> np, vp. Can have a
probability."
pass

class Language:
"A set of rules."

def parse(self, sentence):
"""tries to parse the sentence according to the rules.
returns an iterator over possible parse trees which
are
tuples (category, content). If rules have probability
the iterator returns tuples (parse tree, probability)
in
the order of descending probability."""
pass

def semantics(self, parse):
"returns a semantic interpretation (logic.Clause) of a
given par
pass

class Text:
"A list of sentences who stand in rethorical relations to
each other
pass

class Discourse:
"""A list of texts as utterances of given speakers
(plan.Agent)
which are each supposed to further the established
knowledge about
the discourse universe."""
pass

neural.py

class Neuron:
"Turns a vector of input to an output."
def __init__(self, weights):
self.weights = weights

def __call__(self, input):
"scalar product of input and weights"
pass

class NeuralNetwork:
def __init__(self, input, inter, output):
"""specify size of input matrix, intermediate matrix
and output
neurons will be generated."""
pass

def train(self, input, output):
"train network to produce output from input."
pass

def __call__(self, input):
"produces output matrix from input."
 
G

Gabriel Genellina

I thought about an AI library for python. This is the possible
structure I came up with. Are there any thoughts about it?

Without commenting about the library itself:
None, True and False are spelled this way.
None is a singleton, compare using "x is None" or "x is not None", dont
use ==
This is more stylish, but I prefer to use isxxx() or hasxxx() for
functions that return booleans.
 
R

Roberto Bonvallet

Felix Benner wrote:
[...]
def a-star(self, nodeFrom, nodeTo):
"""searches the shortest path (minimal weight) from
nodeFrom to nodeTo."""
pass
File "<stdin>", line 1
def a-star(self, nodeFrom, nodeTo):
^
SyntaxError: invalid syntax
 
B

bearophileHUGS

Gabriel Genellina:
This is more stylish, but I prefer to use isxxx() or hasxxx() for
functions that return booleans.

Lisp-like languages allow the ending ? or !, I think Ruby allows the
ending ? too (allowing it with Python may be positive). Mathematica
usually uses an ending uppercase Q, like PrimeQ, sometimes I use the
ending Q in Python too.

Bye,
bearophile
 

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question

Members online

No members online now.

Forum statistics

Threads
473,995
Messages
2,570,228
Members
46,818
Latest member
SapanaCarpetStudio

Latest Threads

Top