keithmgould

<script type="text/javascript">
var mpq = [];
mpq.push(["init", "1c8c35d989982401990706fce6659790"]);
(function(){var b,a,e,d,c;b=document.createElement("script");b.type="text/javascript";b.async=true;b.src=(document.location.protocol==="https:"?"https:":"http:")+"//api.mixpanel.com/site_media/js/api/mixpanel.js";a=document.getElementsByTagName("script")[0];a.parentNode.insertBefore(b,a);e=function(f){return function(){mpq.push([f].concat(Array.prototype.slice.call(arguments,0)))}};d=["init","track","track_links","track_forms","register","register_once","identify","name_tag","set_config"];for(c=0;c<d.length;c++){mpq[d[c]]=e(d[c])}})();
</script>

<script type="text/javascript">
var MixPanelHelp = function() {
  return {
    // augments an event with basic info such as current user

keithmgould

define(["constants"], function (constants) {
  var socket = io.connect(constants.socketURL),
      sendables = ["new-message"];

  return {
    listen : function (callback) {
      socket.on("new-message", function (msg) {
        callback({ type : "receive-message", data : msg});
      });
    },

keithmgould

CHAT.namespace("CHAT.CORE");

CHAT.CORE.transport = (function () {
  var socket = io.connect("http://chat.local:3000"),
      sendables = ["new-message"];

  // not happy with the location of this functionality
  socket.on("new-message", function (msg) {
    CHAT.CORE.modules.emit({ type : "receive-message", data : msg});
  });

keithmgould

alias undeployed="heroku releases -a APP_NAME | sed -n 2p | cut -d' ' -f4 | xargs -J % git log --oneline --decorate --color --graph master --not %"

keithmgould

OK. In an effort to learn this stuff (on a Saturday night of course), I also built the tables, indexes, and created fake data.

Here is the big realization: The join is expensive and pointless - it is not needed for the heavy lifting (counting/sorting). Split up into two queries (or a subquery).

First lets do it the original way:

# explain ANALYZE SELECT races.id, races.title, count(participants.id) AS participant_count
FROM races
   INNER JOIN participants ON races.id=participants.race_id

keithmgould

% Taken from Robot
mp=1.793;            % Mass of the pendulum (kg)
mw=0.156;            % Mass of the wheel (kg)
Len=0.209;           % Length to Center of Gravity(meters)
r=0.042;             % Radius of the wheel (meters)
ip=0.07832;          % inertia of the pendulum around the wheel axis
iw=0.000172125;      % inertia of the wheel around the wheel axis

% Taken from Pololu
Rs=0.0024;            % The resistance of the DC Motor

keithmgould

% Declare Constant Values

% Taken from Beaker2 (the robot)
mp=1.793;            % Mass of the pendulum (kg)
mw=0.156;            % Mass of the wheel (kg)
Len=0.209;           % Length to Center of Gravity(meters)
r=0.042;             % Radius of the wheel (meters)
ip=0.07832;          % inertia of the pendulum around the wheel axis
iw=0.000172125;      % inertia of the wheel around the wheel axis

keithmgould

""" 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

keithmgould

import numpy as np

gamma = 0.99

def discount_rewards(r):
    discounted_r = np.zeros_like(r)
    running_add = 0
    for t in reversed(range(0, r.size)):
        running_add = running_add * gamma + r[t]

keithmgould

#!/usr/bin/env python

import gym
import numpy as np
import tensorflow as tf

from tensorflow.python.ops import random_ops

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