reinforcement_learning_for_share_trading\q_learner

q_learner.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import random as rand\n",
    "\n",
    "import numpy as np\n",
    "\n",
    "from indicators import *\n",
    "from col_refs import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "class QLearner(object):\n",
    "\n",
    "    def __init__(self, num_states=100, num_actions=4, alpha=0.2, gamma=0.9, rar=0.98, radr=0.999, dyna=0,\n",
    "                 verbose=False):\n",
    "\n",
    "        self.alpha = alpha\n",
    "        self.gamma = gamma\n",
    "        self.rar = rar\n",
    "        self.radr = radr\n",
    "        self.dyna = dyna\n",
    "        self.verbose = verbose\n",
    "\n",
    "        self.s = 0\n",
    "        self.a = 0\n",
    "\n",
    "        self.Q = np.zeros(shape=(num_states, num_actions))\n",
    "        self.R = np.zeros(shape=(num_states, num_actions))\n",
    "\n",
    "        self.T = np.zeros((num_states, num_actions, num_states))\n",
    "        self.Tc = np.zeros((num_states, num_actions, num_states))\n",
    "\n",
    "        self.num_actions = num_actions\n",
    "        self.num_states = num_states\n",
    "\n",
    "    def querysetstate(self, s):\n",
    "        rand.seed(0)\n",
    "        np.random.seed(0)\n",
    "\n",
    "        if np.random.uniform() < self.rar:\n",
    "            action = rand.randint(0, self.num_actions - 1)\n",
    "\n",
    "        else:\n",
    "            action = self.Q[s, :].argmax()\n",
    "\n",
    "        self.s = s\n",
    "        self.a = action\n",
    "\n",
    "        if self.verbose: print(\"s =\", s, \"a =\", action)\n",
    "\n",
    "        return action\n",
    "\n",
    "    def query(self, s_prime, r):\n",
    "        rand.seed(0)\n",
    "        np.random.seed(0)\n",
    "\n",
    "        if np.random.uniform() < self.rar:\n",
    "            action = rand.randint(0, self.num_actions - 1)\n",
    "\n",
    "        else:\n",
    "            action = self.Q[s_prime, :].argmax()\n",
    "\n",
    "        r_fut = self.Q[s_prime, self.Q[s_prime, :].argmax()]\n",
    "\n",
    "        self.Q[self.s, self.a] = (1 - self.alpha) * self.Q[self.s, self.a] + self.alpha * (r + (self.gamma * r_fut))\n",
    "\n",
    "        self.rar = self.rar * self.radr\n",
    "\n",
    "        if self.dyna != 0:\n",
    "            self.model_update(self.s, self.a, s_prime, r)\n",
    "\n",
    "            self.hallucinate()\n",
    "\n",
    "        self.s = s_prime\n",
    "        self.a = action\n",
    "\n",
    "        if self.verbose: print(\"s =\", s_prime, \"a =\", action, \"r =\", r)\n",
    "\n",
    "        return action\n",
    "\n",
    "    def model_update(self, s, a, s_prime, r):\n",
    "        self.Tc[s, a, s_prime] += 1\n",
    "        self.T = self.Tc / self.Tc.sum(axis=2, keepdims=True)\n",
    "        self.R[s, a] = ((1 - self.alpha) * self.R[s, a]) + (self.alpha * r)\n",
    "\n",
    "    def hallucinate(self):\n",
    "        rand.seed(0)\n",
    "        np.random.seed(0)\n",
    "\n",
    "        for i in range(0, self.dyna):\n",
    "            s_rnd = rand.randint(0, self.num_states - 1)\n",
    "            a_rnd = rand.randint(0, self.num_actions - 1)\n",
    "\n",
    "            s_prime = np.random.multinomial(100, self.T[s_rnd, a_rnd, :]).argmax()\n",
    "\n",
    "            r_rnd = self.R[s_rnd, a_rnd]\n",
    "            r_fut = self.Q[s_prime, self.Q[s_prime, :].argmax()]\n",
    "\n",
    "            self.Q[s_rnd, a_rnd] = (1 - self.alpha) * self.Q[s_rnd, a_rnd] + self.alpha * (r_rnd + (self.gamma * r_fut))\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}