1_part/3_03_RL_Chess_3_q_networks.ipynb

{"nbformat":4,"nbformat_minor":0,"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":"3.6.6"},"colab":{"name":"Reinforcement_Learning_Chess_3_q_networks.ipynb","provenance":[],"collapsed_sections":[],"toc_visible":true}},"cells":[{"cell_type":"markdown","metadata":{"colab_type":"text","id":"hhkArxVLKHF4"},"source":["# 0. 강화학습을 이용한 체스 (Reinforcement Learning Chess) - 3\n","\n","※본 커널은 캐글을 통해 공유된 https://www.kaggle.com/arjanso/reinforcement-learning-chess-3-q-networks 의 내용을 기반으로 작성되었습니다.\n","\n"]},{"cell_type":"markdown","metadata":{"id":"uq_pOluUVeNH","colab_type":"text"},"source":["# Reinforcement Learning Chess\n","Reinforcement Learning Chess는 체스 AI를 개발하기 위해 강화 학습 알고리즘을 구현하는 일련의 Notebook입니다. 간단한 방법으로 다루어 질 수 있는 더 단순한 버전 (환경)으로 시작하여 본격적인 체스 인공 지능이 생길 때까지 이러한 개념을 점차적으로 확장합니다."]},{"cell_type":"markdown","metadata":{"id":"8nOI_En4VeNK","colab_type":"text"},"source":["[**Notebook 1: Policy Iteration**](https://www.kaggle.com/arjanso/reinforcement-learning-chess-1-policy-iteration)  \n","[**Notebook 2: Model-free learning**](https://www.kaggle.com/arjanso/reinforcement-learning-chess-2-model-free-methods)  \n","[**Notebook 4: Policy Gradients**](https://www.kaggle.com/arjanso/reinforcement-learning-chess-4-policy-gradients)  \n","[**Notebook 5: Monte Carlo Tree Search**](https://www.kaggle.com/arjanso/reinforcement-learning-chess-5-tree-search)  "]},{"cell_type":"markdown","metadata":{"id":"BAAn6uenVeNL","colab_type":"text"},"source":["# Notebook III: Q-networks\n","이 notebook에서는 캡처 체스라는 간단한 버전의 체스를 구현합니다. 이 환경(environment)에서 agent(흰색 player)는 말을 잡으면 보상(reward)을 받습니다 (체크 메이트가 아님). 이 notebook을 실행한 후에는, 아래 GIF에 표시되는 것처럼 random oponnent에 대해 말을 캡처할 수 있는 agent를 갖게 됩니다. 이 notebook과 이전 notebook의 주요 차이점은 Q-table 대신 Q-network를 사용한다는 것입니다. Q-table은 훌륭하고 간단하지만 제한된 양의 동작값(action value)만 포함 할 수 있습니다. 체스의 상태(state) 공간 복잡도는 10<sup>47</sup>입니다. 말할 것도 없이, 이것은 Q-table에 넣기에는 너무 많은 정보입니다. 이러한 상황에서 지도 학습을 사용합니다. Q-network는 상태(state)에서 행동값(action value)으로의 일반화된 매핑을 나타낼 수 있습니다.\n","\n","![](https://images.chesscomfiles.com/uploads/game-gifs/90px/green/neo/0/cc/0/0/aXFZUWpyN1Brc1BPbHQwS211WEhudkh6cXohMGFPMExPUTJNUTY4MDY1OTI1NFpSND8yOT85M1Y5MTA3MUxLQ3RDUkpDSjcwTE0wN293V0d6Rzc2cHhWTXJ6NlhzQVg0dUM0WGNNWDU,.gif)\n","\n"]},{"cell_type":"markdown","metadata":{"id":"Ptl__QpHVeNM","colab_type":"text"},"source":["## Import and Install"]},{"cell_type":"code","metadata":{"_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","id":"I0XlcoCjVeNN","colab_type":"code","colab":{}},"source":["import numpy as np # linear algebra\n","import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n","import os\n","import inspect"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","id":"wS6D2JcAVeNS","colab_type":"code","colab":{}},"source":["%load_ext autoreload\n","%autoreload 2"],"execution_count":0,"outputs":[]},{"cell_type":"code","metadata":{"id":"94qnlgjVVeNW","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":216},"outputId":"99b5a3d1-0f0b-44a4-9a74-55bbd3cc8aa4","executionInfo":{"status":"ok","timestamp":1582522509348,"user_tz":-540,"elapsed":9333,"user":{"displayName":"문성채","photoUrl":"","userId":"07687686498757557466"}}},"source":["!pip install python-chess  # Python-Chess is the Python Chess Package that handles the chess environment\n","!pip install --upgrade git+https://github.com/arjangroen/RLC.git  # RLC is the Reinforcement Learning package"],"execution_count":3,"outputs":[{"output_type":"stream","text":["Requirement already satisfied: python-chess in /usr/local/lib/python3.6/dist-packages (0.23.11)\n","Collecting git+https://github.com/arjangroen/RLC.git\n","  Cloning https://github.com/arjangroen/RLC.git to /tmp/pip-req-build-0xc2jmwm\n","  Running command git clone -q https://github.com/arjangroen/RLC.git /tmp/pip-req-build-0xc2jmwm\n","Building wheels for collected packages: RLC\n","  Building wheel for RLC (setup.py) ... \u001b[?25l\u001b[?25hdone\n","  Created wheel for RLC: filename=RLC-0.3-cp36-none-any.whl size=22568 sha256=0e742b043b516bce6fc6ab36a1b9841c1688b79d155d84cfa38b2e465b17868f\n","  Stored in directory: /tmp/pip-ephem-wheel-cache-sippuz9c/wheels/04/68/a5/cb835cd3d76a49de696a942739c71a56bfe66d0d8ea7b4b446\n","Successfully built RLC\n","Installing collected packages: RLC\n","Successfully installed RLC-0.3\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"X1OxbxHlVeNb","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":77},"outputId":"90174eeb-0958-44ab-a107-6b3b67667d1f","executionInfo":{"status":"ok","timestamp":1582522525644,"user_tz":-540,"elapsed":3948,"user":{"displayName":"문성채","photoUrl":"","userId":"07687686498757557466"}}},"source":["import chess\n","from chess.pgn import Game\n","import RLC\n","from RLC.capture_chess.environment import Board\n","from RLC.capture_chess.learn import Q_learning\n","from RLC.capture_chess.agent import Agent"],"execution_count":4,"outputs":[{"output_type":"stream","text":["Using TensorFlow backend.\n"],"name":"stderr"},{"output_type":"display_data","data":{"text/html":["<p style=\"color: red;\">\n","The default version of TensorFlow in Colab will soon switch to TensorFlow 2.x.<br>\n","We recommend you <a href=\"https://www.tensorflow.org/guide/migrate\" target=\"_blank\">upgrade</a> now \n","or ensure your notebook will continue to use TensorFlow 1.x via the <code>%tensorflow_version 1.x</code> magic:\n","<a href=\"https://colab.research.google.com/notebooks/tensorflow_version.ipynb\" target=\"_blank\">more info</a>.</p>\n"],"text/plain":["<IPython.core.display.HTML object>"]},"metadata":{"tags":[]}}]},{"cell_type":"markdown","metadata":{"id":"DZgJWSHTVeNf","colab_type":"text"},"source":["# 3.1 The environment: Capture Chess\n","이 notebook에서는 환경(environment)을 실제 체스처럼 동작하도록 업그레이드 합니다. 주로 python-chess의 Board 객체를 기반으로합니다. 알고리즘이 수렴하기 쉽도록 일부 수정되었습니다.\n","* 환경(environment)이 재설정 된 후, 최대 25번 이동합니다.\n","* agent는 흰색 플레이어 입니다.\n","* 블랙 플레이어는 환경(environment)의 일부이며 무작위 움직임을 반환합니다.\n","* 보상(reward) 구조는 승/패/무승부가 아니라 검은 말을 잡는 것에 따릅니다.\n","    - pawn capture: +1\n","    - knight capture: +3\n","    - bishop capture: +3\n","    - rook capture: +5\n","    - queen capture: +9\n","* 상태(state) 는 8x8x8 배열로 나타냅니다.\n","    - Plane 0 represents pawns\n","    - Plane 1 represents rooks\n","    - Plane 2 represents knights\n","    - Plane 3 represents bishops\n","    - Plane 4 represents queens\n","    - Plane 5 represents kings\n","    - Plane 6 represents 1/총이동횟수(needed for markov property)\n","    - Plane 7 represents can-claim-draw\n","* 흰색 말은 값이 1 이고, 검은색 말은 값이 -1 입니다.\n","       \n"]},{"cell_type":"markdown","metadata":{"id":"dm2Z_aPJVeNg","colab_type":"text"},"source":["## Board representation of python-chess:"]},{"cell_type":"code","metadata":{"id":"MZmiJs_tVeNh","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":421},"outputId":"e987460b-cd23-410e-aa7f-15820d3f030b","executionInfo":{"status":"ok","timestamp":1582522534387,"user_tz":-540,"elapsed":642,"user":{"displayName":"문성채","photoUrl":"","userId":"07687686498757557466"}}},"source":["board = Board()\n","board.board"],"execution_count":5,"outputs":[{"output_type":"execute_result","data":{"text/plain":["Board('rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq - 0 1')"],"image/svg+xml":"<svg height=\"400\" version=\"1.1\" viewBox=\"0 0 400 400\" width=\"400\" xmlns=\"http://www.w3.org/2000/svg\" xmlns:xlink=\"http://www.w3.org/1999/xlink\"><defs><g class=\"white pawn\" id=\"white-pawn\"><path d=\"M22 9c-2.21 0-4 1.79-4 4 0 .89.29 1.71.78 2.38-1.95 1.12-3.28 3.21-3.28 5.62 0 2.03.94 3.84 2.41 5.03-3 1.06-7.41 5.55-7.41 13.47h23c0-7.92-4.41-12.41-7.41-13.47 1.47-1.19 2.41-3 2.41-5.03 0-2.41-1.33-4.5-3.28-5.62.49-.67.78-1.49.78-2.38 0-2.21-1.79-4-4-4z\" fill=\"#fff\" stroke=\"#000\" stroke-linecap=\"round\" stroke-width=\"1.5\" /></g><g class=\"white knight\" fill=\"none\" fill-rule=\"evenodd\" id=\"white-knight\" stroke=\"#000\" stroke-linecap=\"round\" stroke-linejoin=\"round\" stroke-width=\"1.5\"><path d=\"M 22,10 C 32.5,11 38.5,18 38,39 L 15,39 C 15,30 25,32.5 23,18\" style=\"fill:#ffffff; stroke:#000000;\" /><path d=\"M 24,18 C 24.38,20.91 18.45,25.37 16,27 C 13,29 13.18,31.34 11,31 C 9.958,30.06 12.41,27.96 11,28 C 10,28 11.19,29.23 10,30 C 9,30 5.997,31 6,26 C 6,24 12,14 12,14 C 12,14 13.89,12.1 14,10.5 C 13.27,9.506 13.5,8.5 13.5,7.5 C 14.5,6.5 16.5,10 16.5,10 L 18.5,10 C 18.5,10 19.28,8.008 21,7 C 22,7 22,10 22,10\" style=\"fill:#ffffff; stroke:#000000;\" /><path d=\"M 9.5 25.5 A 0.5 0.5 0 1 1 8.5,25.5 A 0.5 0.5 0 1 1 9.5 25.5 z\" style=\"fill:#000000; stroke:#000000;\" /><path d=\"M 15 15.5 A 0.5 1.5 0 1 1 14,15.5 A 0.5 1.5 0 1 1 15 15.5 z\" style=\"fill:#000000; stroke:#000000;\" transform=\"matrix(0.866,0.5,-0.5,0.866,9.693,-5.173)\" /></g><g class=\"white bishop\" fill=\"none\" fill-rule=\"evenodd\" id=\"white-bishop\" stroke=\"#000\" stroke-linecap=\"round\" stroke-linejoin=\"round\" stroke-width=\"1.5\"><g fill=\"#fff\" stroke-linecap=\"butt\"><path d=\"M9 36c3.39-.97 10.11.43 13.5-2 3.39 2.43 10.11 1.03 13.5 2 0 0 1.65.54 3 2-.68.97-1.65.99-3 .5-3.39-.97-10.11.46-13.5-1-3.39 1.46-10.11.03-13.5 1-1.354.49-2.323.47-3-.5 1.354-1.94 3-2 3-2zM15 32c2.5 2.5 12.5 2.5 15 0 .5-1.5 0-2 0-2 0-2.5-2.5-4-2.5-4 5.5-1.5 6-11.5-5-15.5-11 4-10.5 14-5 15.5 0 0-2.5 1.5-2.5 4 0 0-.5.5 0 2zM25 8a2.5 2.5 0 1 1-5 0 2.5 2.5 0 1 1 5 0z\" /></g><path d=\"M17.5 26h10M15 30h15m-7.5-14.5v5M20 18h5\" stroke-linejoin=\"miter\" /></g><g class=\"white rook\" fill=\"#fff\" fill-rule=\"evenodd\" id=\"white-rook\" stroke=\"#000\" stroke-linecap=\"round\" stroke-linejoin=\"round\" stroke-width=\"1.5\"><path d=\"M9 39h27v-3H9v3zM12 36v-4h21v4H12zM11 14V9h4v2h5V9h5v2h5V9h4v5\" stroke-linecap=\"butt\" /><path d=\"M34 14l-3 3H14l-3-3\" /><path d=\"M31 17v12.5H14V17\" stroke-linecap=\"butt\" stroke-linejoin=\"miter\" /><path d=\"M31 29.5l1.5 2.5h-20l1.5-2.5\" /><path d=\"M11 14h23\" fill=\"none\" stroke-linejoin=\"miter\" /></g><g class=\"white queen\" fill=\"#fff\" fill-rule=\"evenodd\" id=\"white-queen\" stroke=\"#000\" stroke-linecap=\"round\" stroke-linejoin=\"round\" stroke-width=\"1.5\"><path d=\"M8 12a2 2 0 1 1-4 0 2 2 0 1 1 4 0zM24.5 7.5a2 2 0 1 1-4 0 2 2 0 1 1 4 0zM41 12a2 2 0 1 1-4 0 2 2 0 1 1 4 0zM16 8.5a2 2 0 1 1-4 0 2 2 0 1 1 4 0zM33 9a2 2 0 1 1-4 0 2 2 0 1 1 4 0z\" /><path d=\"M9 26c8.5-1.5 21-1.5 27 0l2-12-7 11V11l-5.5 13.5-3-15-3 15-5.5-14V25L7 14l2 12zM9 26c0 2 1.5 2 2.5 4 1 1.5 1 1 .5 3.5-1.5 1-1.5 2.5-1.5 2.5-1.5 1.5.5 2.5.5 2.5 6.5 1 16.5 1 23 0 0 0 1.5-1 0-2.5 0 0 .5-1.5-1-2.5-.5-2.5-.5-2 .5-3.5 1-2 2.5-2 2.5-4-8.5-1.5-18.5-1.5-27 0z\" stroke-linecap=\"butt\" /><path d=\"M11.5 30c3.5-1 18.5-1 22 0M12 33.5c6-1 15-1 21 0\" fill=\"none\" /></g><g class=\"white king\" fill=\"none\" fill-rule=\"evenodd\" id=\"white-king\" stroke=\"#000\" stroke-linecap=\"round\" stroke-linejoin=\"round\" stroke-width=\"1.5\"><path d=\"M22.5 11.63V6M20 8h5\" stroke-linejoin=\"miter\" /><path d=\"M22.5 25s4.5-7.5 3-10.5c0 0-1-2.5-3-2.5s-3 2.5-3 2.5c-1.5 3 3 10.5 3 10.5\" fill=\"#fff\" stroke-linecap=\"butt\" stroke-linejoin=\"miter\" /><path d=\"M11.5 37c5.5 3.5 15.5 3.5 21 0v-7s9-4.5 6-10.5c-4-6.5-13.5-3.5-16 4V27v-3.5c-3.5-7.5-13-10.5-16-4-3 6 5 10 5 10V37z\" fill=\"#fff\" /><path d=\"M11.5 30c5.5-3 15.5-3 21 0m-21 3.5c5.5-3 15.5-3 21 0m-21 3.5c5.5-3 15.5-3 21 0\" /></g><g class=\"black pawn\" id=\"black-pawn\"><path d=\"M22 9c-2.21 0-4 1.79-4 4 0 .89.29 1.71.78 2.38-1.95 1.12-3.28 3.21-3.28 5.62 0 2.03.94 3.84 2.41 5.03-3 1.06-7.41 5.55-7.41 13.47h23c0-7.92-4.41-12.41-7.41-13.47 1.47-1.19 2.41-3 2.41-5.03 0-2.41-1.33-4.5-3.28-5.62.49-.67.78-1.49.78-2.38 0-2.21-1.79-4-4-4z\" stroke=\"#000\" stroke-linecap=\"round\" stroke-width=\"1.5\" /></g><g class=\"black knight\" fill=\"none\" fill-rule=\"evenodd\" id=\"black-knight\" stroke=\"#000\" stroke-linecap=\"round\" stroke-linejoin=\"round\" stroke-width=\"1.5\"><path d=\"M 22,10 C 32.5,11 38.5,18 38,39 L 15,39 C 15,30 25,32.5 23,18\" style=\"fill:#000000; stroke:#000000;\" /><path d=\"M 24,18 C 24.38,20.91 18.45,25.37 16,27 C 13,29 13.18,31.34 11,31 C 9.958,30.06 12.41,27.96 11,28 C 10,28 11.19,29.23 10,30 C 9,30 5.997,31 6,26 C 6,24 12,14 12,14 C 12,14 13.89,12.1 14,10.5 C 13.27,9.506 13.5,8.5 13.5,7.5 C 14.5,6.5 16.5,10 16.5,10 L 18.5,10 C 18.5,10 19.28,8.008 21,7 C 22,7 22,10 22,10\" style=\"fill:#000000; stroke:#000000;\" /><path d=\"M 9.5 25.5 A 0.5 0.5 0 1 1 8.5,25.5 A 0.5 0.5 0 1 1 9.5 25.5 z\" style=\"fill:#ececec; stroke:#ececec;\" /><path d=\"M 15 15.5 A 0.5 1.5 0 1 1 14,15.5 A 0.5 1.5 0 1 1 15 15.5 z\" style=\"fill:#ececec; stroke:#ececec;\" transform=\"matrix(0.866,0.5,-0.5,0.866,9.693,-5.173)\" /><path d=\"M 24.55,10.4 L 24.1,11.85 L 24.6,12 C 27.75,13 30.25,14.49 32.5,18.75 C 34.75,23.01 35.75,29.06 35.25,39 L 35.2,39.5 L 37.45,39.5 L 37.5,39 C 38,28.94 36.62,22.15 34.25,17.66 C 31.88,13.17 28.46,11.02 25.06,10.5 L 24.55,10.4 z \" style=\"fill:#ececec; stroke:none;\" /></g><g class=\"black bishop\" fill=\"none\" fill-rule=\"evenodd\" id=\"black-bishop\" stroke=\"#000\" stroke-linecap=\"round\" stroke-linejoin=\"round\" stroke-width=\"1.5\"><path d=\"M9 36c3.39-.97 10.11.43 13.5-2 3.39 2.43 10.11 1.03 13.5 2 0 0 1.65.54 3 2-.68.97-1.65.99-3 .5-3.39-.97-10.11.46-13.5-1-3.39 1.46-10.11.03-13.5 1-1.354.49-2.323.47-3-.5 1.354-1.94 3-2 3-2zm6-4c2.5 2.5 12.5 2.5 15 0 .5-1.5 0-2 0-2 0-2.5-2.5-4-2.5-4 5.5-1.5 6-11.5-5-15.5-11 4-10.5 14-5 15.5 0 0-2.5 1.5-2.5 4 0 0-.5.5 0 2zM25 8a2.5 2.5 0 1 1-5 0 2.5 2.5 0 1 1 5 0z\" fill=\"#000\" stroke-linecap=\"butt\" /><path d=\"M17.5 26h10M15 30h15m-7.5-14.5v5M20 18h5\" stroke=\"#fff\" stroke-linejoin=\"miter\" /></g><g class=\"black rook\" fill=\"#000\" fill-rule=\"evenodd\" id=\"black-rook\" stroke=\"#000\" stroke-linecap=\"round\" stroke-linejoin=\"round\" stroke-width=\"1.5\"><path d=\"M9 39h27v-3H9v3zM12.5 32l1.5-2.5h17l1.5 2.5h-20zM12 36v-4h21v4H12z\" stroke-linecap=\"butt\" /><path d=\"M14 29.5v-13h17v13H14z\" stroke-linecap=\"butt\" stroke-linejoin=\"miter\" /><path d=\"M14 16.5L11 14h23l-3 2.5H14zM11 14V9h4v2h5V9h5v2h5V9h4v5H11z\" stroke-linecap=\"butt\" /><path d=\"M12 35.5h21M13 31.5h19M14 29.5h17M14 16.5h17M11 14h23\" fill=\"none\" stroke=\"#fff\" stroke-linejoin=\"miter\" stroke-width=\"1\" /></g><g class=\"black queen\" fill=\"#000\" fill-rule=\"evenodd\" id=\"black-queen\" stroke=\"#000\" stroke-linecap=\"round\" stroke-linejoin=\"round\" stroke-width=\"1.5\"><g fill=\"#000\" stroke=\"none\"><circle cx=\"6\" cy=\"12\" r=\"2.75\" /><circle cx=\"14\" cy=\"9\" r=\"2.75\" /><circle cx=\"22.5\" cy=\"8\" r=\"2.75\" /><circle cx=\"31\" cy=\"9\" r=\"2.75\" /><circle cx=\"39\" cy=\"12\" r=\"2.75\" /></g><path d=\"M9 26c8.5-1.5 21-1.5 27 0l2.5-12.5L31 25l-.3-14.1-5.2 13.6-3-14.5-3 14.5-5.2-13.6L14 25 6.5 13.5 9 26zM9 26c0 2 1.5 2 2.5 4 1 1.5 1 1 .5 3.5-1.5 1-1.5 2.5-1.5 2.5-1.5 1.5.5 2.5.5 2.5 6.5 1 16.5 1 23 0 0 0 1.5-1 0-2.5 0 0 .5-1.5-1-2.5-.5-2.5-.5-2 .5-3.5 1-2 2.5-2 2.5-4-8.5-1.5-18.5-1.5-27 0z\" stroke-linecap=\"butt\" /><path d=\"M11 38.5a35 35 1 0 0 23 0\" fill=\"none\" stroke-linecap=\"butt\" /><path d=\"M11 29a35 35 1 0 1 23 0M12.5 31.5h20M11.5 34.5a35 35 1 0 0 22 0M10.5 37.5a35 35 1 0 0 24 0\" fill=\"none\" stroke=\"#fff\" /></g><g class=\"black king\" fill=\"none\" fill-rule=\"evenodd\" id=\"black-king\" stroke=\"#000\" stroke-linecap=\"round\" stroke-linejoin=\"round\" stroke-width=\"1.5\"><path d=\"M22.5 11.63V6\" stroke-linejoin=\"miter\" /><path d=\"M22.5 25s4.5-7.5 3-10.5c0 0-1-2.5-3-2.5s-3 2.5-3 2.5c-1.5 3 3 10.5 3 10.5\" fill=\"#000\" stroke-linecap=\"butt\" stroke-linejoin=\"miter\" /><path d=\"M11.5 37c5.5 3.5 15.5 3.5 21 0v-7s9-4.5 6-10.5c-4-6.5-13.5-3.5-16 4V27v-3.5c-3.5-7.5-13-10.5-16-4-3 6 5 10 5 10V37z\" fill=\"#000\" /><path d=\"M20 8h5\" stroke-linejoin=\"miter\" /><path d=\"M32 29.5s8.5-4 6.03-9.65C34.15 14 25 18 22.5 24.5l.01 2.1-.01-2.1C20 18 9.906 14 6.997 19.85c-2.497 5.65 4.853 9 4.853 9M11.5 30c5.5-3 15.5-3 21 0m-21 3.5c5.5-3 15.5-3 21 0m-21 3.5c5.5-3 15.5-3 21 0\" stroke=\"#fff\" /></g></defs><rect class=\"square dark a1\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"20\" y=\"335\" /><use transform=\"translate(20, 335)\" xlink:href=\"#white-rook\" /><rect class=\"square light b1\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"65\" y=\"335\" /><use transform=\"translate(65, 335)\" xlink:href=\"#white-knight\" /><rect class=\"square dark c1\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"110\" y=\"335\" /><use transform=\"translate(110, 335)\" xlink:href=\"#white-bishop\" /><rect class=\"square light d1\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"155\" y=\"335\" /><use transform=\"translate(155, 335)\" xlink:href=\"#white-queen\" /><rect class=\"square dark e1\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"200\" y=\"335\" /><use transform=\"translate(200, 335)\" xlink:href=\"#white-king\" /><rect class=\"square light f1\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"245\" y=\"335\" /><use transform=\"translate(245, 335)\" xlink:href=\"#white-bishop\" /><rect class=\"square dark g1\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"290\" y=\"335\" /><use transform=\"translate(290, 335)\" xlink:href=\"#white-knight\" /><rect class=\"square light h1\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"335\" y=\"335\" /><use transform=\"translate(335, 335)\" xlink:href=\"#white-rook\" /><rect class=\"square light a2\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"20\" y=\"290\" /><use transform=\"translate(20, 290)\" xlink:href=\"#white-pawn\" /><rect class=\"square dark b2\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"65\" y=\"290\" /><use transform=\"translate(65, 290)\" xlink:href=\"#white-pawn\" /><rect class=\"square light c2\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"110\" y=\"290\" /><use transform=\"translate(110, 290)\" xlink:href=\"#white-pawn\" /><rect class=\"square dark d2\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"155\" y=\"290\" /><use transform=\"translate(155, 290)\" xlink:href=\"#white-pawn\" /><rect class=\"square light e2\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"200\" y=\"290\" /><use transform=\"translate(200, 290)\" xlink:href=\"#white-pawn\" /><rect class=\"square dark f2\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"245\" y=\"290\" /><use transform=\"translate(245, 290)\" xlink:href=\"#white-pawn\" /><rect class=\"square light g2\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"290\" y=\"290\" /><use transform=\"translate(290, 290)\" xlink:href=\"#white-pawn\" /><rect class=\"square dark h2\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"335\" y=\"290\" /><use transform=\"translate(335, 290)\" xlink:href=\"#white-pawn\" /><rect class=\"square dark a3\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"20\" y=\"245\" /><rect class=\"square light b3\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"65\" y=\"245\" /><rect class=\"square dark c3\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"110\" y=\"245\" /><rect class=\"square light d3\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"155\" y=\"245\" /><rect class=\"square dark e3\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"200\" y=\"245\" /><rect class=\"square light f3\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"245\" y=\"245\" /><rect class=\"square dark g3\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"290\" y=\"245\" /><rect class=\"square light h3\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"335\" y=\"245\" /><rect class=\"square light a4\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"20\" y=\"200\" /><rect class=\"square dark b4\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"65\" y=\"200\" /><rect class=\"square light c4\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"110\" y=\"200\" /><rect class=\"square dark d4\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"155\" y=\"200\" /><rect class=\"square light e4\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"200\" y=\"200\" /><rect class=\"square dark f4\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"245\" y=\"200\" /><rect class=\"square light g4\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"290\" y=\"200\" /><rect class=\"square dark h4\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"335\" y=\"200\" /><rect class=\"square dark a5\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"20\" y=\"155\" /><rect class=\"square light b5\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"65\" y=\"155\" /><rect class=\"square dark c5\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"110\" y=\"155\" /><rect class=\"square light d5\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"155\" y=\"155\" /><rect class=\"square dark e5\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"200\" y=\"155\" /><rect class=\"square light f5\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"245\" y=\"155\" /><rect class=\"square dark g5\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"290\" y=\"155\" /><rect class=\"square light h5\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"335\" y=\"155\" /><rect class=\"square light a6\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"20\" y=\"110\" /><rect class=\"square dark b6\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"65\" y=\"110\" /><rect class=\"square light c6\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"110\" y=\"110\" /><rect class=\"square dark d6\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"155\" y=\"110\" /><rect class=\"square light e6\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"200\" y=\"110\" /><rect class=\"square dark f6\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"245\" y=\"110\" /><rect class=\"square light g6\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"290\" y=\"110\" /><rect class=\"square dark h6\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"335\" y=\"110\" /><rect class=\"square dark a7\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"20\" y=\"65\" /><use transform=\"translate(20, 65)\" xlink:href=\"#black-pawn\" /><rect class=\"square light b7\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"65\" y=\"65\" /><use transform=\"translate(65, 65)\" xlink:href=\"#black-pawn\" /><rect class=\"square dark c7\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"110\" y=\"65\" /><use transform=\"translate(110, 65)\" xlink:href=\"#black-pawn\" /><rect class=\"square light d7\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"155\" y=\"65\" /><use transform=\"translate(155, 65)\" xlink:href=\"#black-pawn\" /><rect class=\"square dark e7\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"200\" y=\"65\" /><use transform=\"translate(200, 65)\" xlink:href=\"#black-pawn\" /><rect class=\"square light f7\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"245\" y=\"65\" /><use transform=\"translate(245, 65)\" xlink:href=\"#black-pawn\" /><rect class=\"square dark g7\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"290\" y=\"65\" /><use transform=\"translate(290, 65)\" xlink:href=\"#black-pawn\" /><rect class=\"square light h7\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"335\" y=\"65\" /><use transform=\"translate(335, 65)\" xlink:href=\"#black-pawn\" /><rect class=\"square light a8\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"20\" y=\"20\" /><use transform=\"translate(20, 20)\" xlink:href=\"#black-rook\" /><rect class=\"square dark b8\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"65\" y=\"20\" /><use transform=\"translate(65, 20)\" xlink:href=\"#black-knight\" /><rect class=\"square light c8\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"110\" y=\"20\" /><use transform=\"translate(110, 20)\" xlink:href=\"#black-bishop\" /><rect class=\"square dark d8\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"155\" y=\"20\" /><use transform=\"translate(155, 20)\" xlink:href=\"#black-queen\" /><rect class=\"square light e8\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"200\" y=\"20\" /><use transform=\"translate(200, 20)\" xlink:href=\"#black-king\" /><rect class=\"square dark f8\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"245\" y=\"20\" /><use transform=\"translate(245, 20)\" xlink:href=\"#black-bishop\" /><rect class=\"square light g8\" fill=\"#ffce9e\" height=\"45\" stroke=\"none\" width=\"45\" x=\"290\" y=\"20\" /><use transform=\"translate(290, 20)\" xlink:href=\"#black-knight\" /><rect class=\"square dark h8\" fill=\"#d18b47\" height=\"45\" stroke=\"none\" width=\"45\" x=\"335\" y=\"20\" /><use transform=\"translate(335, 20)\" xlink:href=\"#black-rook\" /><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"42\" y=\"10\">a</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"42\" y=\"390\">a</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"87\" y=\"10\">b</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"87\" y=\"390\">b</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"132\" y=\"10\">c</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"132\" y=\"390\">c</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"177\" y=\"10\">d</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"177\" y=\"390\">d</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"222\" y=\"10\">e</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"222\" y=\"390\">e</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"267\" y=\"10\">f</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"267\" y=\"390\">f</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"312\" y=\"10\">g</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"312\" y=\"390\">g</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"357\" y=\"10\">h</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"357\" y=\"390\">h</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"10\" y=\"357\">1</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"390\" y=\"357\">1</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"10\" y=\"312\">2</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"390\" y=\"312\">2</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"10\" y=\"267\">3</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"390\" y=\"267\">3</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"10\" y=\"222\">4</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"390\" y=\"222\">4</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"10\" y=\"177\">5</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"390\" y=\"177\">5</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"10\" y=\"132\">6</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"390\" y=\"132\">6</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"10\" y=\"87\">7</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"390\" y=\"87\">7</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"10\" y=\"42\">8</text><text alignment-baseline=\"middle\" font-size=\"14\" text-anchor=\"middle\" x=\"390\" y=\"42\">8</text></svg>"},"metadata":{"tags":[]},"execution_count":5}]},{"cell_type":"markdown","metadata":{"id":"Kf_7bZVNVeNm","colab_type":"text"},"source":["### Numerical representation of the pawns (layer 0)\n","다른 말을 보려면 첫번째 dimension 의 index 를 변경하십시요.\n","\n","[0,::-1,:] 에서 0은 pawn 을 지정한 것 입니다."]},{"cell_type":"code","metadata":{"id":"1N5bwIJNVeNn","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":147},"outputId":"87a30656-cd0d-4399-a7b8-717385ab3329","executionInfo":{"status":"ok","timestamp":1582522576183,"user_tz":-540,"elapsed":634,"user":{"displayName":"문성채","photoUrl":"","userId":"07687686498757557466"}}},"source":["board.layer_board[0,::-1,:].astype(int)"],"execution_count":11,"outputs":[{"output_type":"execute_result","data":{"text/plain":["array([[ 0,  0,  0,  0,  0,  0,  0,  0],\n","       [-1, -1, -1, -1, -1, -1, -1, -1],\n","       [ 0,  0,  0,  0,  0,  0,  0,  0],\n","       [ 0,  0,  0,  0,  0,  0,  0,  0],\n","       [ 0,  0,  0,  0,  0,  0,  0,  0],\n","       [ 0,  0,  0,  0,  0,  0,  0,  0],\n","       [ 1,  1,  1,  1,  1,  1,  1,  1],\n","       [ 0,  0,  0,  0,  0,  0,  0,  0]])"]},"metadata":{"tags":[]},"execution_count":11}]},{"cell_type":"markdown","metadata":{"id":"auEFouT0VeNq","colab_type":"text"},"source":["## The Agent\n","* agent 는 더 이상 하나의 말이 아니라 체스 플레이어입니다.\n","* action space는 64x64 = 4096 동작(action)으로 구성됩니다.\n","  * 말을 집어 올릴 수 있는 8x8 = 64 곳이 있습니다\n","  * 그리고 말을 내려 놓을 수 있는 또 다른 64 곳이 있습니다.\n","* 물론 특정 action만 합법적입니다. 특정 상태(state)에서 어떤 action 이 합법적인지는 환경(environment)의 일부입니다. (RL에서는 agent의 통제를 벗어난 모든 것이 환경의 일부로 간주 됨). python-chess 패키지를 사용하여 합법적 인 움직임을 선택할 수 있습니다. (AlphaZero는 유사한 접근 방식을 사용하는 것 같습니다.\n","https://ai.stackexchange.com/questions/7979/why-does-the-policy-network-in-alphazero-work)"]},{"cell_type":"markdown","metadata":{"id":"do-gsIa-VeNr","colab_type":"text"},"source":["### Implementation"]},{"cell_type":"code","metadata":{"id":"Zyjid8GNVeNs","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":736},"outputId":"ebf05997-660f-4d25-e1d4-52740cf948fa","executionInfo":{"status":"ok","timestamp":1582522542635,"user_tz":-540,"elapsed":1143,"user":{"displayName":"문성채","photoUrl":"","userId":"07687686498757557466"}}},"source":["board = Board()\n","agent = Agent(network='conv',gamma=0.1,lr=0.07)\n","R = Q_learning(agent,board)\n","R.agent.fix_model()\n","R.agent.model.summary()"],"execution_count":7,"outputs":[{"output_type":"stream","text":["WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:541: The name tf.placeholder is deprecated. Please use tf.compat.v1.placeholder instead.\n","\n","WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:66: The name tf.get_default_graph is deprecated. Please use tf.compat.v1.get_default_graph instead.\n","\n","WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:4432: The name tf.random_uniform is deprecated. Please use tf.random.uniform instead.\n","\n","WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:190: The name tf.get_default_session is deprecated. Please use tf.compat.v1.get_default_session instead.\n","\n","WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:197: The name tf.ConfigProto is deprecated. Please use tf.compat.v1.ConfigProto instead.\n","\n","WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:203: The name tf.Session is deprecated. Please use tf.compat.v1.Session instead.\n","\n","WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:207: The name tf.global_variables is deprecated. Please use tf.compat.v1.global_variables instead.\n","\n","WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:216: The name tf.is_variable_initialized is deprecated. Please use tf.compat.v1.is_variable_initialized instead.\n","\n","WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:223: The name tf.variables_initializer is deprecated. Please use tf.compat.v1.variables_initializer instead.\n","\n","WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/optimizers.py:793: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.\n","\n","Model: \"model_1\"\n","__________________________________________________________________________________________________\n","Layer (type)                    Output Shape         Param #     Connected to                     \n","==================================================================================================\n","board_layer (InputLayer)        (None, 8, 8, 8)      0                                            \n","__________________________________________________________________________________________________\n","conv2d_1 (Conv2D)               (None, 1, 8, 8)      9           board_layer[0][0]                \n","__________________________________________________________________________________________________\n","conv2d_2 (Conv2D)               (None, 1, 8, 8)      9           board_layer[0][0]                \n","__________________________________________________________________________________________________\n","reshape_1 (Reshape)             (None, 1, 64)        0           conv2d_1[0][0]                   \n","__________________________________________________________________________________________________\n","reshape_2 (Reshape)             (None, 1, 64)        0           conv2d_2[0][0]                   \n","__________________________________________________________________________________________________\n","dot_1 (Dot)                     (None, 64, 64)       0           reshape_1[0][0]                  \n","                                                                 reshape_2[0][0]                  \n","__________________________________________________________________________________________________\n","reshape_3 (Reshape)             (None, 4096)         0           dot_1[0][0]                      \n","==================================================================================================\n","Total params: 18\n","Trainable params: 18\n","Non-trainable params: 0\n","__________________________________________________________________________________________________\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"les47IzJVeNw","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":797},"outputId":"2cd9574f-b63c-427b-e163-3fa25cfee434","executionInfo":{"status":"ok","timestamp":1582522546731,"user_tz":-540,"elapsed":712,"user":{"displayName":"문성채","photoUrl":"","userId":"07687686498757557466"}}},"source":["print(inspect.getsource(agent.network_update))"],"execution_count":8,"outputs":[{"output_type":"stream","text":["    def network_update(self, minibatch):\n","        \"\"\"\n","        Update the Q-network using samples from the minibatch\n","        Args:\n","            minibatch: list\n","                The minibatch contains the states, moves, rewards and new states.\n","\n","        Returns:\n","            td_errors: np.array\n","                array of temporal difference errors\n","\n","        \"\"\"\n","\n","        # Prepare separate lists\n","        states, moves, rewards, new_states = [], [], [], []\n","        td_errors = []\n","        episode_ends = []\n","        for sample in minibatch:\n","            states.append(sample[0])\n","            moves.append(sample[1])\n","            rewards.append(sample[2])\n","            new_states.append(sample[3])\n","\n","            # Episode end detection\n","            if np.array_equal(sample[3], sample[3] * 0):\n","                episode_ends.append(0)\n","            else:\n","                episode_ends.append(1)\n","\n","        # The Q target\n","        q_target = np.array(rewards) + np.array(episode_ends) * self.gamma * np.max(\n","            self.fixed_model.predict(np.stack(new_states, axis=0)), axis=1)\n","\n","        # The Q value for the remaining actions\n","        q_state = self.model.predict(np.stack(states, axis=0))  # batch x 64 x 64\n","\n","        # Combine the Q target with the other Q values.\n","        q_state = np.reshape(q_state, (len(minibatch), 64, 64))\n","        for idx, move in enumerate(moves):\n","            td_errors.append(q_state[idx, move[0], move[1]] - q_target[idx])\n","            q_state[idx, move[0], move[1]] = q_target[idx]\n","        q_state = np.reshape(q_state, (len(minibatch), 4096))\n","\n","        # Perform a step of minibatch Gradient Descent.\n","        self.model.fit(x=np.stack(states, axis=0), y=q_state, epochs=1, verbose=0)\n","\n","        return td_errors\n","\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"HirnIY7CVeNz","colab_type":"text"},"source":["### Q learning with a Q-network\n","**Theory**\n","- Q-network 는 일반적으로 선형회귀(linear regression) 또는 (deep) neural network입니다.\n","- 네트워크의 입력은 상태(state) (S)이고 출력은 각 동작(action)의 예상 동작값(action value) 입니다.(여기서는 4096 개의 값)\n","- 아이디어는 Q-table을 사용한 학습과 유사합니다. 할인 보상값(discounted reward) 과 최대 후속 상태 동작값(max successor state action value) 를 합한 값의 방향으로 Q 값을 업데이트 합니다.\n","- 우선 순위가 지정된 경험 재생을 사용하여 업데이트의 상관 관계를 제거했습니다. 더 자세히 알고 싶다면 참고 문헌의 링크를 확인하십시오.\n","  - 학습 과정을 안정화시키기 위해 fixed-Q 목표를 사용했습니다."]},{"cell_type":"markdown","metadata":{"id":"G7JREs1GVeN0","colab_type":"text"},"source":["### Implementation\n","- linear network 와 convolutional network, 두 개의 network를 구축\n","- linear model 은 상태(state) (8,8,8)를 동작(action) (64,64)에 매핑하여 32k가 넘는 훈련 가능한 가중치를 제공합니다! 매개 변수 공유가 없기 때문에 매우 비효율적이지만 작동합니다.\n","- Convolutional model 은 2개의 1x1 convulutions를 사용하고 결과 배열의 outer product을 취합니다. 결과적으로 훈련 가능한 가중치는 18 개뿐입니다!\n","  - 장점 : 더 많은 매개 변수 공유-> 더 빠른 수렴\n","  - 단점 : 정보 손실-> 성능 저하\n","- 실제 체스 AI에는 더 큰 신경망이 필요합니다. 그러나 이제 신경망은 값어치 있는 말을 잡는 법을 배워야합니다."]},{"cell_type":"code","metadata":{"id":"eePFmJuvVeN1","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":1000},"outputId":"55e48457-3ab5-4fa7-d53a-a5ed2bae34ff","executionInfo":{"status":"ok","timestamp":1582522552478,"user_tz":-540,"elapsed":834,"user":{"displayName":"문성채","photoUrl":"","userId":"07687686498757557466"}}},"source":["print(inspect.getsource(R.play_game))"],"execution_count":9,"outputs":[{"output_type":"stream","text":["    def play_game(self, k, greedy=False, maxiter=25):\n","        \"\"\"\n","        Play a game of capture chess\n","        Args:\n","            k: int\n","                game count, determines epsilon (exploration rate)\n","            greedy: Boolean\n","                if greedy, no exploration is done\n","            maxiter: int\n","                Maximum amount of steps per game\n","\n","        Returns:\n","\n","        \"\"\"\n","        episode_end = False\n","        turncount = 0\n","\n","        # Here we determine the exploration rate. k is divided by 250 to slow down the exploration rate decay.\n","        eps = max(0.05, 1 / (1 + (k / 250))) if not greedy else 0.\n","\n","        # Play a game of chess\n","        while not episode_end:\n","            state = self.env.layer_board\n","            explore = np.random.uniform(0, 1) < eps  # determine whether to explore\n","            if explore:\n","                move = self.env.get_random_action()\n","                move_from = move.from_square\n","                move_to = move.to_square\n","            else:\n","                action_values = self.agent.get_action_values(np.expand_dims(state, axis=0))\n","                action_values = np.reshape(np.squeeze(action_values), (64, 64))\n","                action_space = self.env.project_legal_moves()  # The environment determines which moves are legal\n","                action_values = np.multiply(action_values, action_space)\n","                move_from = np.argmax(action_values, axis=None) // 64\n","                move_to = np.argmax(action_values, axis=None) % 64\n","                moves = [x for x in self.env.board.generate_legal_moves() if \\\n","                         x.from_square == move_from and x.to_square == move_to]\n","                if len(moves) == 0:  # If all legal moves have negative action value, explore.\n","                    move = self.env.get_random_action()\n","                    move_from = move.from_square\n","                    move_to = move.to_square\n","                else:\n","                    move = np.random.choice(moves)  # If there are multiple max-moves, pick a random one.\n","\n","            episode_end, reward = self.env.step(move)\n","            new_state = self.env.layer_board\n","            if len(self.memory) > self.memsize:\n","                self.memory.pop(0)\n","                self.sampling_probs.pop(0)\n","            turncount += 1\n","            if turncount > maxiter:\n","                episode_end = True\n","                reward = 0\n","            if episode_end:\n","                new_state = new_state * 0\n","            self.memory.append([state, (move_from, move_to), reward, new_state])\n","            self.sampling_probs.append(1)\n","\n","            self.reward_trace.append(reward)\n","\n","            self.update_agent(turncount)\n","\n","        return self.env.board\n","\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"Ol9XQIh3VeN4","colab_type":"text"},"source":["### Demo"]},{"cell_type":"code","metadata":{"id":"cnQtLBI4VeN5","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":1000},"outputId":"8e33db08-9b7d-4d04-c90b-be42170f1d1a","executionInfo":{"status":"ok","timestamp":1582528482617,"user_tz":-540,"elapsed":5433048,"user":{"displayName":"문성채","photoUrl":"","userId":"07687686498757557466"}}},"source":["pgn = R.learn(iters=750)"],"execution_count":12,"outputs":[{"output_type":"stream","text":["iter 0\n","WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:1033: The name tf.assign_add is deprecated. Please use tf.compat.v1.assign_add instead.\n","\n","WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:1020: The name tf.assign is deprecated. Please use tf.compat.v1.assign instead.\n","\n","iter 10\n","iter 20\n","iter 30\n","iter 40\n","iter 50\n","iter 60\n","iter 70\n","iter 80\n","iter 90\n","iter 100\n","iter 110\n","iter 120\n","iter 130\n","iter 140\n","iter 150\n","iter 160\n","iter 170\n","iter 180\n","iter 190\n","iter 200\n","iter 210\n","iter 220\n","iter 230\n","iter 240\n","iter 250\n","iter 260\n","iter 270\n","iter 280\n","iter 290\n","iter 300\n","iter 310\n","iter 320\n","iter 330\n","iter 340\n","iter 350\n","iter 360\n","iter 370\n","iter 380\n","iter 390\n","iter 400\n","iter 410\n","iter 420\n","iter 430\n","iter 440\n","iter 450\n","iter 460\n","iter 470\n","iter 480\n","iter 490\n","iter 500\n","iter 510\n","iter 520\n","iter 530\n","iter 540\n","iter 550\n","iter 560\n","iter 570\n","iter 580\n","iter 590\n","iter 600\n","iter 610\n","iter 620\n","iter 630\n","iter 640\n","iter 650\n","iter 660\n","iter 670\n","iter 680\n","iter 690\n","iter 700\n","iter 710\n","iter 720\n","iter 730\n","iter 740\n"],"name":"stdout"},{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAXQAAAD4CAYAAAD8Zh1EAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"tags":[]}}]},{"cell_type":"code","metadata":{"id":"7l4XcHQ5VeN9","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":568},"outputId":"ece7cffd-e1ac-4760-e0c7-c3cc5f0e0830","executionInfo":{"status":"ok","timestamp":1582530359186,"user_tz":-540,"elapsed":1253,"user":{"displayName":"문성채","photoUrl":"","userId":"07687686498757557466"}}},"source":["reward_smooth = pd.DataFrame(R.reward_trace)\n","reward_smooth.rolling(window=125,min_periods=0).mean().plot(figsize=(16,9),title='average performance over the last 125 steps')"],"execution_count":13,"outputs":[{"output_type":"execute_result","data":{"text/plain":["<matplotlib.axes._subplots.AxesSubplot at 0x7f2b1a1629e8>"]},"metadata":{"tags":[]},"execution_count":13},{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAA6IAAAIYCAYAAAB33lEgAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRK5CYII=\n","text/plain":["<Figure size 1152x648 with 1 Axes>"]},"metadata":{"tags":[]}}]},{"cell_type":"markdown","metadata":{"id":"rKxypexQVeOB","colab_type":"text"},"source":["PGN 파일이 출력 폴더로 내보내집니다. [chess.com analysis board](https://www.chess.com/analysis) 에 붙여 넣어 분석할 수 있습니다"]},{"cell_type":"code","metadata":{"id":"TV-q7YFPVeOC","colab_type":"code","colab":{}},"source":["with open(\"final_game.pgn\",\"w\") as log:\n","    log.write(str(pgn))"],"execution_count":0,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"gsGuIi9lVeOH","colab_type":"text"},"source":["# 3.2 Learned action values analysis\n","네트워크는 무엇을 배웠습니까? 아래 코드는 모든 흰색말에 대해 모든 검은색 말을 잡는 동작값(action value)을 확인합니다.\n","- 검은색 말을 잡기 위한 동작값(action value)이 환경(environment)에 설정한 (Reinfeld) 보상(reward)과 비슷할 것으로 기대합니다.\n","- 물론 동작값(action value)은 블랙 플레이어에 의해서 다시 잡히는 위험과 연속으로 말을 잡는 기회에 따라 달라집니다."]},{"cell_type":"code","metadata":{"id":"h3LmYZPNVeOI","colab_type":"code","colab":{}},"source":["board.reset()\n","bl = board.layer_board\n","bl[6,:,:] = 1/10  # Assume we are in move 10\n","av = R.agent.get_action_values(np.expand_dims(bl,axis=0))\n","\n","av = av.reshape((64,64))\n","\n","p = board.board.piece_at(20)#.symbol()\n","\n","\n","white_pieces = ['P','N','B','R','Q','K']\n","black_piece = ['_','p','n','b','r','q','k']\n","\n","df = pd.DataFrame(np.zeros((6,7)))\n","\n","df.index = white_pieces\n","df.columns = black_piece\n","\n","for from_square in range(16):\n","    for to_square in range(30,64):\n","        from_piece = board.board.piece_at(from_square).symbol()\n","        to_piece = board.board.piece_at(to_square)\n","        if to_piece:\n","            to_piece = to_piece.symbol()\n","        else:\n","            to_piece = '_'\n","        df.loc[from_piece,to_piece] = av[from_square,to_square]\n","        \n","        "],"execution_count":0,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"IXduqnEjVeOL","colab_type":"text"},"source":["* ###흰색(대문자) 말로 검은색(소문자) 말을 잡기 위한 학습된 동작값(action value)입니다. 밑줄은 아무 것도 못 잡는 것을 나타냅니다."]},{"cell_type":"code","metadata":{"id":"H90_ML2PVeOL","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":220},"outputId":"1997282c-bbb9-4f94-b9a8-bbaab72025cb","executionInfo":{"status":"ok","timestamp":1582530372365,"user_tz":-540,"elapsed":596,"user":{"displayName":"문성채","photoUrl":"","userId":"07687686498757557466"}}},"source":["df[['_','p','n','b','r','q']]"],"execution_count":16,"outputs":[{"output_type":"execute_result","data":{"text/html":["<div>\n","<style scoped>\n","    .dataframe tbody tr th:only-of-type {\n","        vertical-align: middle;\n","    }\n","\n","    .dataframe tbody tr th {\n","        vertical-align: top;\n","    }\n","\n","    .dataframe thead th {\n","        text-align: right;\n","    }\n","</style>\n","<table border=\"1\" class=\"dataframe\">\n","  <thead>\n","    <tr style=\"text-align: right;\">\n","      <th></th>\n","      <th>_</th>\n","      <th>p</th>\n","      <th>n</th>\n","      <th>b</th>\n","      <th>r</th>\n","      <th>q</th>\n","    </tr>\n","  </thead>\n","  <tbody>\n","    <tr>\n","      <th>P</th>\n","      <td>0.798019</td>\n","      <td>1.563570</td>\n","      <td>3.540396</td>\n","      <td>3.993215</td>\n","      <td>6.070866</td>\n","      <td>9.135963</td>\n","    </tr>\n","    <tr>\n","      <th>N</th>\n","      <td>0.720169</td>\n","      <td>1.411037</td>\n","      <td>3.195016</td>\n","      <td>3.603660</td>\n","      <td>5.478627</td>\n","      <td>8.244712</td>\n","    </tr>\n","    <tr>\n","      <th>B</th>\n","      <td>0.737288</td>\n","      <td>1.444579</td>\n","      <td>3.270964</td>\n","      <td>3.689322</td>\n","      <td>5.608859</td>\n","      <td>8.440696</td>\n","    </tr>\n","    <tr>\n","      <th>R</th>\n","      <td>0.673799</td>\n","      <td>1.320183</td>\n","      <td>2.989294</td>\n","      <td>3.371626</td>\n","      <td>5.125867</td>\n","      <td>7.713848</td>\n","    </tr>\n","    <tr>\n","      <th>Q</th>\n","      <td>0.661777</td>\n","      <td>1.296630</td>\n","      <td>2.935962</td>\n","      <td>3.311474</td>\n","      <td>5.034418</td>\n","      <td>7.576227</td>\n","    </tr>\n","    <tr>\n","      <th>K</th>\n","      <td>0.666056</td>\n","      <td>1.305012</td>\n","      <td>2.954942</td>\n","      <td>3.332881</td>\n","      <td>5.066963</td>\n","      <td>7.625205</td>\n","    </tr>\n","  </tbody>\n","</table>\n","</div>"],"text/plain":["          _         p         n         b         r         q\n","P  0.798019  1.563570  3.540396  3.993215  6.070866  9.135963\n","N  0.720169  1.411037  3.195016  3.603660  5.478627  8.244712\n","B  0.737288  1.444579  3.270964  3.689322  5.608859  8.440696\n","R  0.673799  1.320183  2.989294  3.371626  5.125867  7.713848\n","Q  0.661777  1.296630  2.935962  3.311474  5.034418  7.576227\n","K  0.666056  1.305012  2.954942  3.332881  5.066963  7.625205"]},"metadata":{"tags":[]},"execution_count":16}]},{"cell_type":"markdown","metadata":{"id":"cDAB4UlWVeOO","colab_type":"text"},"source":["# References\n","Reinforcement Learning: An Introduction  \n","> Richard S. Sutton and Andrew G. Barto  \n","> 1st Edition  \n","> MIT Press, march 1998  \n","\n","RL Course by David Silver: Lecture playlist  \n","> https://www.youtube.com/watch?v=2pWv7GOvuf0&list=PLqYmG7hTraZDM-OYHWgPebj2MfCFzFObQ  \n","\n","Experience Replay  \n","> https://datascience.stackexchange.com/questions/20535/what-is-experience-replay-and-what-are-its-benefits"]}]}