ematvey

import math
from typing import Union, NamedTuple

import numpy as np
from scipy.ndimage import affine_transform


class AffineTransform(NamedTuple):
    matrix: np.ndarray
    offset: np.ndarray

louisng5

I have used some aggressive measure to achieve 19 Episodes before solve, inculding:
1) Large batch_size used for experience replay.
2) 30 epoch size used for every training batch.
3) 0 probability for random action.

With this approach, the NN start to "know" how to balance at episode 10-15.

Keras was used for creating and training the NN.

karpathy

""" Trains an agent with (stochastic) Policy Gradients on Pong. Uses OpenAI Gym. """
import numpy as np
import cPickle as pickle
import gym

# hyperparameters
H = 200 # number of hidden layer neurons
batch_size = 10 # every how many episodes to do a param update?
learning_rate = 1e-4
gamma = 0.99 # discount factor for reward

martinrusev

[supervisord]
nodaemon=true
loglevel=debug

[program:amon]
command=gunicorn -c gunicorn.conf.py wsgi
directory=/amon
autostart=true
autorestart=true
redirect_stderr=true 

bellbind

"""Genetic Algorithmn Implementation
see:
http://www.obitko.com/tutorials/genetic-algorithms/ga-basic-description.php
"""
import random

class GeneticAlgorithm(object):
    def __init__(self, genetics):
        self.genetics = genetics
        pass