keithmgould

# toy example from https://www.kdnuggets.com/2017/04/simple-understand-gradient-descent-algorithm.html
import numpy as np
import random

import matplotlib.pyplot as plt

LEARNING_RATE_M = 1e-8
LEARNING_RATE_B = 1e-2

# sq ft , price

keithmgould

0 - rwd:269.0, loss:-14.074104309082031
1 - rwd:206.0, loss:7.945431232452393
2 - rwd:236.0, loss:-2.8392391204833984
3 - rwd:228.75, loss:-41.76856231689453
4 - rwd:250.8, loss:-11.660385131835938
5 - rwd:231.5, loss:11.052635192871094
6 - rwd:217.14285714285714, loss:5.223728179931641
7 - rwd:214.5, loss:-6.977025985717773
8 - rwd:205.88888888888889, loss:-6.304763317108154
9 - rwd:207.5, loss:-2.8461992740631104

keithmgould

import numpy as np
from keras.models import Sequential
from keras.callbacks import TensorBoard
from keras.layers import *
from keras.optimizers import *

N=10
INPUT_LEN = 4
OUTPUT_LEN = 2

keithmgould

# OpenGym MountainCar-v0
# -------------------
#
# This code demonstrates debugging of a basic Q-network (without target network)
# in an OpenGym MountainCar-v0 environment.
#
# Made as part of blog series Let's make a DQN, available at: 
# https://jaromiru.com/2016/10/12/lets-make-a-dqn-debugging/
# 
# author: Jaromir Janisch, 2016

keithmgould

conda install -c anaconda numpy scipy pillow mpi4py
conda install -c conda-forge matplotlib
pip install tensorflow keras gym jupyter

keithmgould

<?xml version="1.0"?>
<robot name="bot">

  <!-- colors ................................................................. -->
  <material name="red">
    <color rgba="0.8 0 0 1"/>
  </material>

 <material name="green">
    <color rgba="0 0.8 0 0.75"/>

keithmgould

[user]
	name = Keith Gould
	email = keithmgould@gmail.com
[core]
	excludesfile = ~/.gitignore_global
[credential]
	helper = osxkeychain
[alias]
  b = branch
  c = commit

keithmgould

"------------------------------------
" Look and Feel
"------------------------------------
set guifont=Monaco:h11
set background=dark

" show line numbers
set number
set ruler

keithmgould

import pdb
import gym
import numpy as np
import tensorflow as tf
import matplotlib.pyplot as plt

from tensorflow.python.ops import random_ops

def _initializer(shape, dtype=tf.float32, partition_info=None):
     return random_ops.random_normal(shape)

keithmgould

import numpy as np
import datetime
import tensorflow as tf

class Policy():
    def __init__(self, session):
        self.session = session
        self.actions = tf.placeholder(tf.float32)
        self.advantages = tf.placeholder(tf.float32)
        self.observations = tf.placeholder(tf.float32, shape=[None, 4])