Deep Learning with Shogi AI on Mac and Google Colab Chapter 11

TOP PAGE

Learning technique

Ordinary learning

Learning results of the value network in Chapter 10. image.png

Transfer learning

Learning by transferring the learning results of the policy network. The error was low and the accuracy was high. image.png

Multitask learning

The upper row is Chapter 7 Policy Network The bottom row is Chapter 10 Value Network The idea of multitask learning is to make it common because the light blue parts are the same. image.png This is what happens when the light blue is standardized. image.png

You can learn policies and values at the same time. The accuracy is also good. image.png

Residual Network It seems that the configuration called ResNet is good. I don't understand why ResNet is good, but it seems that research is progressing.

1 block of ResNet image.png

Details of one block of ResNet image.png

Connect 5 ResNet blocks and replace them with L2 to L12 for multitask learning. image.png

Learning results Learning is progressing faster than without ResNet. image.png

policy_value_resnet.py

x + h2 x + h2 is the operation of adding x and h2 "element by element". (When I actually printed the value, it was like that)

    def __call__(self, x):
        h1 = F.relu(self.bn1(self.conv1(x)))
        h2 = self.bn2(self.conv2(h1))
        return F.relu(x + h2)

h = self['b{}'.format(i)](h) This way of writing means self.bi (h).

        for i in range(1, self.blocks + 1):
            h = self['b{}'.format(i)](h) 

pydlshogi/network/policy_value_resnet.py


#!/usr/bin/env python3
# -*- coding: utf-8 -*-

from chainer import Chain
import chainer.functions as F
import chainer.links as L

from pydlshogi.common import *

ch = 192
fcl = 256

class Block(Chain):
    def __init__(self):
        super(Block, self).__init__()
        with self.init_scope():
            self.conv1 = L.Convolution2D(in_channels = ch, out_channels = ch, ksize = 3, pad = 1, nobias=True)
            self.bn1   = L.BatchNormalization(ch)
            self.conv2 = L.Convolution2D(in_channels = ch, out_channels = ch, ksize = 3, pad = 1, nobias=True)
            self.bn2   = L.BatchNormalization(ch)

    def __call__(self, x):
        h1 = F.relu(self.bn1(self.conv1(x)))
        h2 = self.bn2(self.conv2(h1))
        return F.relu(x + h2)
        # x +When h2 printed the value and confirmed it, x and h2 were added element by element. In other words, x, h2, x + h2 are 194 elements.
        #I was worried that the first feeling would be 388 elements, so I checked it, but F.Does it mean that each element is added in the relu?

class PolicyValueResnet(Chain):
    def __init__(self, blocks):
        super(PolicyValueResnet, self).__init__()
        self.blocks = blocks
        with self.init_scope():
            self.l1 = L.Convolution2D(in_channels = 104, out_channels = ch, ksize = 3, pad = 1)
            for i in range(1, blocks + 1):
                self.add_link('b{}'.format(i), Block()) #The first argument is the name and the second argument is the class
            # policy network
            self.lpolicy = L.Convolution2D(in_channels = ch, out_channels = MOVE_DIRECTION_LABEL_NUM, ksize = 1, nobias = True)
            self.lpolicy_bias = L.Bias(shape=(9*9*MOVE_DIRECTION_LABEL_NUM))
            # value network
            self.lvalue1 = L.Convolution2D(in_channels = ch, out_channels = MOVE_DIRECTION_LABEL_NUM, ksize = 1)
            self.lvalue2 = L.Linear(9*9*MOVE_DIRECTION_LABEL_NUM, fcl)
            self.lvalue3 = L.Linear(fcl, 1)

    def __call__(self, x):
        h = F.relu(self.l1(x))
        for i in range(1, self.blocks + 1):
            h = self['b{}'.format(i)](h) #This way of writing self.b How many.
        # policy network
        h_policy = self.lpolicy(h)
        policy = self.lpolicy_bias(F.reshape(h_policy, (-1, 9*9*MOVE_DIRECTION_LABEL_NUM)))
        # value network
        h_value = F.relu(self.lvalue1(h))
        h_value = F.relu(self.lvalue2(h_value))
        value = self.lvalue3(h_value)
        return policy, value

Recommended Posts

Deep Learning with Shogi AI on Mac and Google Colab Chapter 11
Deep Learning with Shogi AI on Mac and Google Colab Chapter 8
Deep Learning with Shogi AI on Mac and Google Colab Chapter 12 3
Deep Learning with Shogi AI on Mac and Google Colab Chapter 7
Deep Learning with Shogi AI on Mac and Google Colab Chapter 10 6-9
Deep Learning with Shogi AI on Mac and Google Colab Chapter 10
Deep Learning with Shogi AI on Mac and Google Colab Chapter 7 5-7
Deep Learning with Shogi AI on Mac and Google Colab Chapter 9
Deep Learning with Shogi AI on Mac and Google Colab Chapter 12 3
Deep Learning with Shogi AI on Mac and Google Colab Chapter 12 3
Deep Learning with Shogi AI on Mac and Google Colab Chapter 12 1-2
Deep Learning with Shogi AI on Mac and Google Colab Chapter 12 3
Deep Learning with Shogi AI on Mac and Google Colab Chapter 12 3 ~ 5
Deep Learning with Shogi AI on Mac and Google Colab Chapter 7 9
Deep Learning with Shogi AI on Mac and Google Colab Chapter 8 5-9
Deep Learning with Shogi AI on Mac and Google Colab Chapter 8 1-4
Deep Learning with Shogi AI on Mac and Google Colab Chapter 12 3
Deep Learning with Shogi AI on Mac and Google Colab Chapter 7 8
Deep Learning with Shogi AI on Mac and Google Colab Chapter 7 1-4
Deep Learning with Shogi AI on Mac and Google Colab
Deep Learning with Shogi AI on Mac and Google Colab Chapters 1-6
Learn with Shogi AI Deep Learning on Mac and Google Colab Use Google Colab
Deep Learning on Mac and Google Colab Words Learned with Shogi AI
Machine learning with Pytorch on Google Colab
About learning with google colab
Steps to quickly create a deep learning environment on Mac with TensorFlow and OpenCV
Play with Turtle on Google Colab
Use MeCab and neologd with Google Colab
"Learning word2vec" and "Visualization with Tensorboard" on Colaboratory
Deep Learning from scratch The theory and implementation of deep learning learned with Python Chapter 3
Install selenium on Mac and try it with python
Deep learning image analysis starting with Kaggle and Keras
[AI] Deep Metric Learning
Extract music features with Deep Learning and predict tags
"Deep Learning from scratch" Self-study memo (No. 14) Run the program in Chapter 4 on Google Colaboratory
[Google Colab] How to interrupt learning and then resume it
Recognize your boss and hide the screen with Deep Learning
An error that stumbled upon learning YOLO on Google Colab
Machine learning environment settings based on Python 3 on Mac (coexistence with Python 2)
HIKAKIN and Max Murai with live game video and deep learning
Easy deep learning web app with NNC and Python + Flask
Try deep learning with TensorFlow
Deep Kernel Learning with Pyro
Plotly Dash on Google Colab
Try Deep Learning with FPGA
Catalina on Mac and pyenv
Generate Pokemon with Deep Learning
Create AtCoder Contest appointments on Google Calendar with Python and GAS
Build a Python environment on your Mac with Anaconda and PyCharm
Error and solution when installing python3 with homebrew on mac (catalina 10.15)
How to run Jupyter and Spark on Mac with minimal settings
The strongest way to use MeCab and CaboCha with Google Colab
[Reading Notes] Hands-on Machine Learning with Scikit-Learn, Keras, and TensorFlow Chapter 1
Install lp_solve on Mac OS X and call it with python.
Deep Learning / Deep Learning from Zero 2 Chapter 4 Memo
Try Deep Learning with FPGA-Select Cucumbers
Cat breed identification with deep learning
Deep Learning / Deep Learning from Zero Chapter 3 Memo
tensor flow with anaconda on mac
MQTT on Raspberry Pi and Mac
Make ASCII art with deep learning