Last active
April 11, 2020 05:29
-
-
Save jeacom25b/879c59eb0b72ca9b0dbe3442628f7579 to your computer and use it in GitHub Desktop.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
from random import randint, shuffle, choice, random | |
from collections import defaultdict | |
from math import log10 | |
import time | |
import re | |
import os | |
SEQ_LENGTH = 10 | |
POPULATION = 100 | |
REPETITION_PENALITY = 0.9 | |
INPUT_TEXT = "data/gaming.txt" | |
RULES = ( | |
(2, 0.001), | |
(3, 0.01), | |
(4, 0.1), | |
(5, 1), | |
(6, 10), | |
(7, 100), | |
(8, -100000), | |
) | |
TARGET_WORDS = () | |
def pick(vocabulary, fequency_factor=20): | |
index = round((random() ** fequency_factor) * (len(vocabulary) - 1)) | |
return vocabulary[index] | |
def pick2(vocabulary, fequency_factor=8): | |
return pick(vocabulary, fequency_factor), pick(vocabulary, fequency_factor) | |
def neighborhood(seq, idx, radius): | |
if radius + idx <= len(seq): | |
return tuple(seq[idx:idx + radius]) | |
def score_word_seq(word_seq, snippets_frequency): | |
score = 0 | |
multiplier = 1 | |
seen = set() | |
for word in word_seq: | |
if word in seen: | |
multiplier *= REPETITION_PENALITY | |
else: | |
seen.add(word) | |
for target_word, target_count in TARGET_WORDS: | |
if word_seq.count(target_word) != target_count: | |
multiplier *= REPETITION_PENALITY | |
for radius, score_multiplier in RULES: | |
for i in range(len(word_seq)): | |
neigh = neighborhood(word_seq, i, radius) | |
score += score_multiplier if snippets_frequency.get(neigh) else 0 | |
return score * multiplier | |
def mutate(word_seq, context, vocabulary, memory, word_swap=0.03, split_swap=0.03): | |
word_seq = word_seq.copy() | |
index0 = randint(0, len(word_seq) - 1) | |
index1 = randint(0, len(word_seq) - 1) | |
index2 = randint(0, len(word_seq) - 1) | |
word_seq[index1] = pick(vocabulary) | |
if random() < word_swap: | |
word = choice(memory) | |
if word in context: | |
word_seq[index1] = choice(context[word]) | |
if random() < word_swap: | |
word_seq[index2] = pick(memory) | |
if random() < word_swap: | |
word_seq[index0], word_seq[index1] = word_seq[index1], word_seq[index0] | |
if random() < split_swap: | |
word_seq = word_seq[index0:] + word_seq[:index0] | |
return word_seq | |
def crosover(word_seq_a, word_seq_b): | |
return list(choice(x) for x in zip(word_seq_a, word_seq_b)) | |
def context_gen(words_seq): | |
context = defaultdict(set) | |
for i in range(len(words_seq)): | |
word_a = words_seq[i] | |
context[word_a] |= set(words_seq[i - 5:i]) | |
for key in context: | |
context[key] = list(context[key]) | |
return context | |
def preprocess(text): | |
text = text.lower().replace("\n", " ") | |
for i in range(5): | |
text = text.replace(" ", " ") | |
return text.split(" ") | |
vocabulary = [] | |
context = {} | |
snippets_frequency = {} | |
with open(INPUT_TEXT, "r") as file: | |
string = file.read() | |
words_seq = preprocess(string) | |
print(words_seq) | |
print("so many words") | |
print("Bulding a huge datastructure, hol-don!\n\n\n") | |
context = context_gen(words_seq) | |
words_frequency = {} | |
for word in words_seq: | |
words_frequency.setdefault(word, 0) | |
words_frequency[word] += 1 | |
for key in words_frequency: | |
words_frequency[key] = log10(words_frequency[key]) | |
for radius, score_multiplier in RULES: | |
for i in range(len(words_seq)): | |
neigh = neighborhood(words_seq, i, radius) | |
if neigh: | |
snippets_frequency.setdefault(neigh, 0) | |
snippets_frequency[neigh] += 1 | |
vocabulary = list(item[0] for item in sorted(words_frequency.items(), key=lambda i: i[1], reverse=True)) | |
vocabulary = list(word for word, count in TARGET_WORDS) + vocabulary | |
def score(indiv): | |
if not done: | |
return score_word_seq(indiv, snippets_frequency) | |
else: | |
return score_word_seq(done[-1] + indiv, snippets_frequency) | |
individuals = [[pick(vocabulary) for _ in range(SEQ_LENGTH)] for _ in range(POPULATION)] | |
step = 0 | |
done = [] | |
memory = ["what"] | |
while True: | |
step += 1 | |
individuals = list(sorted(individuals, key=score, reverse=True)) | |
if step % 100 == 0: | |
os.system("clear") | |
print("\n".join(" ".join(indiv) for indiv in individuals[:5])) | |
print("\n") | |
print("top gene score: ", score(individuals[0])) | |
print(f"generation: {step}") | |
print("memory: ", memory) | |
print() | |
print(" ".join(" ".join(indiv) for indiv in done)) | |
if step % 600 == 0: | |
step = 0 | |
done.append(individuals[0]) | |
memory.extend(individuals[0]) | |
memory = list(sorted(set(memory), key=lambda w: words_frequency.get(w, -1)))[:5] | |
individuals = [[pick(vocabulary) for _ in range(SEQ_LENGTH)] for _ in range(POPULATION)] | |
individuals = [individuals[0]] + [mutate(crosover(*pick2(individuals)), context, vocabulary, memory) for _ in range(POPULATION - 1)] |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment