一起学习用Verilog在FPGA上实现CNN----(四)池化层设计

news2024/10/5 17:26:03

1 池化层设计

自顶而下分析池化层的设计过程

1.1 Average Pool Multi Layer

图为该项目的平均池化层，其包含一个AvgPoolSingle单元，模块的输入为图像特征矩阵，输出为池化后的特征矩阵

在这里插入图片描述
图片来自附带的技术文档《Hardware Documentation》

池化层的原理图如图所示，其中输入位宽为75264，输出位宽为18816。池化层位于卷积层和激活层之后，第一次卷积层输出位宽为75264，因此池化层的输入位宽为75264。Average Pool Multi Layer的深度为6，前卷积层的输出特征H和W均为28，故输入位宽为28x28x6x16=75264；平均池化窗口大小为2x2，输出特征H和W变为14，输出位宽为14x14x6x16=18816

在这里插入图片描述

1.2 Average Pool Single Layer

该单元是执行单个通道的平均池化操作，由多个AvgU组成，如图所示：

在这里插入图片描述
图片来自附带的技术文档《Hardware Documentation》

原理图如图所示，该单元输入位宽为12544，输出位宽为3136。前一层卷积输出特征H为28，W为28，Average Pool Single Layer的深度为1，因此输入位宽为28x28x1x16=12544；该项目平均池化的窗口大小为2x2，故输出位宽为14x14x1x16=3136

在这里插入图片描述

1.3 Averaging Unit

Averaging Unit如图所示，求输入4个数的均值。该单元先求4个数A、B、C、D的和，再将和乘以0.25得到4个数的均值

在这里插入图片描述
图片来自附带的技术文档《Hardware Documentation》

原理图如图所示，输入为4个位宽16的数，输出为位宽16的均值

在这里插入图片描述

2 代码实现

2.1 AvgUnit

2.1.1 设计输入

创建AvgUnit文件，操作如图：

在这里插入图片描述

输入文件名：

在这里插入图片描述

确认创建：

在这里插入图片描述

双击打开，输入代码：

module AvgUnit (numA,numB,numC,numD,AvgOut);
  
parameter DATA_WIDTH = 16;

input [DATA_WIDTH-1:0] numA,numB,numC,numD;
output [DATA_WIDTH-1:0] AvgOut;

wire [DATA_WIDTH-1:0] add1result;
wire [DATA_WIDTH-1:0] add2result;
wire [DATA_WIDTH-1:0] add3result;
reg [DATA_WIDTH-1:0] quarter = 16'b0011010000000000;

floatAdd16 FADD1 (numA,numB,add1result);
floatAdd16 FADD2 (add1result,numC,add2result);
floatAdd16 FADD3 (add2result,numD,add3result);  
floatMult16 FM (add3result,quarter,AvgOut);  
endmodule

如图所示：

在这里插入图片描述

2.1.2 分析与综合

将AvgUnit设置为顶层：

在这里插入图片描述

对设计进行分析，操作如图：

在这里插入图片描述

分析后的设计，Vivado自动生成原理图，如图：

在这里插入图片描述

AvgUnit原理图如图：

在这里插入图片描述

对设计进行综合，操作如图：

在这里插入图片描述

综合完成，关闭即可：

在这里插入图片描述

2.1.3 功能仿真

创建TestBench，操作如图所示：

在这里插入图片描述

输入激励文件名tb_AvgUnit：

在这里插入图片描述

双击打开，输入激励代码：

`timescale 1ns / 1ps
module tb_AvgUnit();
reg [15:0] numA;
reg [15:0] numB;
reg [15:0] numC;
reg [15:0] numD;
wire [15:0] AvgOut;

initial begin
	
	// Avg(2,3,4,5) = 3.5
	#0
  numA = 16'h4000;
  numB = 16'h4200;
  numC = 16'h4400;
  numD = 16'h4500;

	// Avg(1,2,3,4) = 2.5
	#10
	numA = 16'h3C00;
  numB = 16'h4000;
  numC = 16'h4200;
  numD = 16'h4400;
	
	// Avg(-1,-2,-3,-4) = -2.5
	#10
	numA = 16'hBC00;
  numB = 16'hC000;
  numC = 16'hC200;
  numD = 16'hC400;
  
  // Avg(-2,-3,-4,-5) = 3.5
  #10
	numA = 16'hC000;
  numB = 16'hC200;
  numC = 16'hC400;
  numD = 16'hC500;
  #10
	$stop;
end

AvgUnit UUT
(
    .numA(numA),
    .numB(numB),
    .numC(numC),
    .numD(numD),
    .AvgOut(AvgOut)
);

endmodule

如图所示：

在这里插入图片描述

将tb_AvgUnit设置为顶层：

在这里插入图片描述

开始进行仿真，操作如下:

在这里插入图片描述

开始仿真，如图：

在这里插入图片描述

仿真波形，如图所示：

在这里插入图片描述

仿真结束，关闭仿真：

在这里插入图片描述

2.2 AvgPoolSingle

2.2.1 设计输入

创建AvgPoolSingle文件，如图：

在这里插入图片描述

双击打开，输入代码：

module AvgPoolSingle(aPoolIn,aPoolOut);
  
parameter DATA_WIDTH = 16;
parameter InputH = 28;
parameter InputW = 28;
parameter Depth = 1;

input [0:InputH*InputW*Depth*DATA_WIDTH-1] aPoolIn;
output [0:(InputH/2)*(InputW/2)*Depth*DATA_WIDTH-1] aPoolOut;

genvar i,j;

generate  // 2x2的池化窗口 
  for (i=0; i<(InputH); i=i+2) begin
    for (j=0; j<(InputW); j=j+2) begin
    AvgUnit
    #(
     .DATA_WIDTH(DATA_WIDTH)
     )
     AU
    (
      .numA(aPoolIn[(i*InputH+j)*DATA_WIDTH+:DATA_WIDTH]),
      .numB(aPoolIn[(i*InputH+j+1)*DATA_WIDTH+:DATA_WIDTH]),
      .numC(aPoolIn[((i+1)*InputH+j)*DATA_WIDTH+:DATA_WIDTH]),
      .numD(aPoolIn[((i+1)*InputH+j+1)*DATA_WIDTH+:DATA_WIDTH]),
      .AvgOut(aPoolOut[(i/2*InputH/2+j/2)*DATA_WIDTH+:DATA_WIDTH])
      );
    end
  end
endgenerate
endmodule

如图所示：

在这里插入图片描述

2.2.2 分析与综合

将AvgPoolSingle文件设置为顶层：

在这里插入图片描述

关闭上次的分析文件：

在这里插入图片描述

对本次设计进行分析，操作如图：

在这里插入图片描述

分析后的设计，Vivado自动生成原理图，如图：

在这里插入图片描述

该模块由多个AvgUnit组成，如图：

在这里插入图片描述

对设计进行综合，操作如图：

在这里插入图片描述

综合完成，关闭即可：

在这里插入图片描述

2.2.3 功能仿真

创建TestBench，操作如图所示：

在这里插入图片描述

双击打开，输入激励代码：

`timescale 1ns / 1ps
module tb_AvgPoolSingle();
reg [28*28*16-1:0] inAvg;
wire [14*14*16-1:0] outAvg;

initial begin
    #0
    inAvg = 12544'h4000440040004400400044004000440040004400400044004000440040004400400044004000440040004400400044004000440040004400450042004500420045004200450042004500420045004200450042004500420045004200450042004500420045004200450042004500420040004400400044004000440040004400400044004000440040004400400044004000440040004400400044004000440040004400400044004500420045004200450042004500420045004200450042004500420045004200450042004500420045004200450042004500420045004200400044004000440040004400400044004000440040004400400044004000440040004400400044004000440040004400400044004000440045004200450042004500420045004200450042004500420045004200450042004500420045004200450042004500420045004200450042004000440040004400400044004000440040004400400044004000440040004400400044004000440040004400400044004000440040004400450042004500420045004200450042004500420045004200450042004500420045004200450042004500420045004200450042004500420040004400400044004000440040004400400044004000440040004400400044004000440040004400400044004000440040004400400044004500420045004200450042004500420045004200450042004500420045004200450042004500420045004200450042004500420045004200400044004000440040004400400044004000440040004400400044004000440040004400400044004000440040004400400044004000440045004200450042004500420045004200450042004500420045004200450042004500420045004200450042004500420045004200450042004000440040004400400044004000440040004400400044004000440040004400400044004000440040004400400044004000440040004400450042004500420045004200450042004500420045004200450042004500420045004200450042004500420045004200450042004500420040004400400044004000440040004400400044004000440040004400400044004000440040004400400044004000440040004400400044004500420045004200450042004500420045004200450042004500420045004200450042004500420045004200450042004500420045004200400044004000440040004400400044004000440040004400400044004000440040004400400044004000440040004400400044004000440045004200450042004500420045004200450042004500420045004200450042004500420045004200450042004500420045004200450042004000440040004400400044004000440040004400400044004000440040004400400044004000440040004400400044004000440040004400450042004500420045004200450042004500420045004200450042004500420045004200450042004500420045004200450042004500420040004400400044004000440040004400400044004000440040004400400044004000440040004400400044004000440040004400400044004500420045004200450042004500420045004200450042004500420045004200450042004500420045004200450042004500420045004200400044004000440040004400400044004000440040004400400044004000440040004400400044004000440040004400400044004000440045004200450042004500420045004200450042004500420045004200450042004500420045004200450042004500420045004200450042004000440040004400400044004000440040004400400044004000440040004400400044004000440040004400400044004000440040004400450042004500420045004200450042004500420045004200450042004500420045004200450042004500420045004200450042004500420040004400400044004000440040004400400044004000440040004400400044004000440040004400400044004000440040004400400044004500420045004200450042004500420045004200450042004500420045004200450042004500420045004200450042004500420045004200;
    #10
    $stop;
end
AvgPoolSingle UUT
  (
    .aPoolIn(inAvg),
    .aPoolOut(outAvg)
  );
endmodule

如图所示：

在这里插入图片描述
将tb_AvgPoolSingle设置为顶层：

在这里插入图片描述

开始进行仿真，操作如图：

在这里插入图片描述

如图，开始仿真：

在这里插入图片描述

仿真波形，如图：

在这里插入图片描述

仿真完成，关闭仿真：

在这里插入图片描述

2.3 AvgPoolMulti

2.3.1 设计输入

创建AvgPoolMulti文件，如图：

在这里插入图片描述

双击打开，输入代码：

module AvgPoolMulti(clk, reset, apInput, apOutput);

parameter DATA_WIDTH = 16;
parameter D = 6;
parameter H = 28;
parameter W = 28;

input reset,clk;
input [0:H*W*D*DATA_WIDTH-1] apInput;
output reg [0:(H/2)*(W/2)*D*DATA_WIDTH-1] apOutput;

reg [0:H*W*DATA_WIDTH-1] apInput_s;
wire [0:(H/2)*(W/2)*DATA_WIDTH-1] apOutput_s;
integer counter;


AvgPoolSingle
  #(
      .DATA_WIDTH(DATA_WIDTH),
      .InputH(H),
      .InputW(W)
  ) AvgPool
  (
      .aPoolIn(apInput_s),
      .aPoolOut(apOutput_s)
  );

always @ (posedge clk or posedge reset) begin
  if (reset == 1'b1) begin
    counter = 0;
  end
  else if (counter<D) begin 
    counter = counter+1;
  end
end

always @ (*) begin
  apInput_s = apInput[counter*H*W*DATA_WIDTH+:H*W*DATA_WIDTH];
  apOutput[counter*(H/2)*(W/2)*DATA_WIDTH+:(H/2)*(W/2)*DATA_WIDTH] = apOutput_s;
end

endmodule

如图所示：

在这里插入图片描述

将AvgPoolMulti设置为顶层：

在这里插入图片描述

2.3.2 分析与综合

对设计进行分析，操作如图：

在这里插入图片描述

分析后的设计，Vivado自动生成原理图，如图：

在这里插入图片描述

原理图如图所示：

在这里插入图片描述

对设计进行综合，操作如图：

在这里插入图片描述

2.3.3 功能仿真

创建TestBench，操作如图所示：

在这里插入图片描述
双击打开，输入激励代码：

`timescale 1ns / 1ps
module tb_AvgPoolMulti();
reg clk,reset;
reg [6*28*28*16-1:0] inAvg;
wire [6*14*14*16-1:0] outAvg;

localparam PERIOD = 100;

integer i;

always
	#(PERIOD/2) clk = ~clk;
	
initial begin
    #0
    clk = 1'b0;
	  reset = 1;
    inAvg = 75264'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;
    #(PERIOD)
	  reset = 0;
    #(8*PERIOD)
    for (i = 6*14*14-1; i >=0; i = i - 1) begin
		  $displayh(outAvg[i*16+:16]);
	  end
    $stop;
end
AvgPoolMulti UUT
  (
    .clk(clk),
    .reset(reset),
    .apInput(inAvg),
    .apOutput(outAvg)
  );
endmodule

如图所示：

在这里插入图片描述
将tb_AvgPoolMulti设置为顶层：

在这里插入图片描述

开始进行仿真，操作如下:

在这里插入图片描述

如图，开始仿真：

在这里插入图片描述

仿真波形，如图所示：

在这里插入图片描述

仿真结束，关闭仿真：

在这里插入图片描述

2.4 IntegrationConv

2.4.1 设计输入

双击打开integrationConv文件，修改代码为：

module integrationConv (clk,reset,CNNinput,Conv1F,Conv2F,iConvOutput);

parameter DATA_WIDTH = 16;
parameter ImgInW = 32;
parameter ImgInH = 32;
parameter Conv1Out = 28;
parameter AvgP1out = 14;
parameter Conv2Out = 10;
parameter Kernel = 5;
parameter AvgP2out = 5;
parameter DepthC1 = 6;
parameter DepthC2 = 16;


input clk, reset;
input [ImgInW*ImgInH*DATA_WIDTH-1:0] CNNinput;
input [Kernel*Kernel*DepthC1*DATA_WIDTH-1:0] Conv1F;
input [DepthC2*Kernel*Kernel*DepthC1*DATA_WIDTH-1:0] Conv2F;
output [DepthC2*DATA_WIDTH-1:0] iConvOutput;

reg C1rst,AP1rst,Tanh1Reset;
wire Tanh1Flag;

wire [Conv1Out*Conv1Out*DepthC1*DATA_WIDTH-1:0] C1out;
wire [Conv1Out*Conv1Out*DepthC1*DATA_WIDTH-1:0] C1outTanH;

wire [AvgP1out*AvgP1out*DepthC1*DATA_WIDTH-1:0] AP1out;

convLayerMulti C1
(
	.clk(clk),
	.reset(reset),
	.image(CNNinput),
	.filters(Conv1F),
	.outputConv(C1out)
);

UsingTheTanh16
#(.nofinputs(Conv1Out*Conv1Out*DepthC1))
Tanh1(
      .x(C1out),
      .clk(clk),
      .Output(C1outTanH),
      .resetExternal(Tanh1Reset),
      .FinishedTanh(Tanh1Flag)
      );

AvgPoolMulti AP1
  (
    .clk(clk),
    .reset(reset),
    .apInput(C1outTanH),
    .apOutput(AP1out)
  );
convLayerMulti
#(
  .DATA_WIDTH(16),
  .D(6),
  .H(14),
  .W(14),
  .F(5),
  .K(16)
) C2 
(
	.clk(clk),
	.reset(reset),
	.image(AP1out),
	.filters(Conv2F),
	.outputConv(iConvOutput)
);
endmodule

如图所示：

在这里插入图片描述

将integrationConv设置为顶层：

在这里插入图片描述

关闭上次的分析文件：

在这里插入图片描述

2.4.2 分析与综合

对设计进行分析，操作如图：

在这里插入图片描述

分析后的设计，Vivado自动生成原理图，如图：

在这里插入图片描述

顶层原理图，如图所示：

在这里插入图片描述

对设计进行综合，操作如图：

在这里插入图片描述

希望本文对大家有帮助，上文若有不妥之处，欢迎指正

分享决定高度，学习拉开差距

本文来自互联网用户投稿，该文观点仅代表作者本人，不代表本站立场。本站仅提供信息存储空间服务，不拥有所有权，不承担相关法律责任。如若转载，请注明出处：http://www.coloradmin.cn/o/137211.html

如若内容造成侵权/违法违规/事实不符，请联系多彩编程网进行投诉反馈，一经查实，立即删除！

相关文章

Java开发学习(三十六)----SpringBoot三种配置文件解析

Java开发学习(三十六)----SpringBoot三种配置文件解析

一、 配置文件格式 我们现在启动服务器默认的端口号是 8080&#xff0c;访问路径可以书写为 http://localhost:8080/books/1 在线上环境我们还是希望将端口号改为 80&#xff0c;这样在访问的时候就可以不写端口号了&#xff0c;如下 http://localhost/books/1 而 SpringB…
阅读更多...
电脑屏录软件，这3款良心软件，分享给你

电脑屏录软件，这3款良心软件，分享给你

现在很多人会使用电脑屏录软件&#xff0c;有些用来记录游戏中的精彩操作&#xff0c;有些用来记录在线教学课程&#xff0c;有些用来记录在线视频会议。现在有各种各样的电脑屏录软件。选择一个好的电脑屏录软件是非常重要的。电脑屏录软件哪个好&#xff1f;下面小编分享3款良…
阅读更多...
前端基础(四)_数据类型的强制转换

前端基础(四)_数据类型的强制转换

数据类型的强制转换就是通过js提供的函数进行数据转换。常见的就是将其他类型的数据转换成number类型和string类型。 一、其他类型转 number 类型 1.Number Number 方法将其他类型的数据转换为Number类型&#xff0c;返回一个新的数值&#xff0c;不会改变变量本身。 例1&…
阅读更多...
【Linux】Linux编辑器-vim的使用以及指令集

【Linux】Linux编辑器-vim的使用以及指令集

推荐先将vim配置好后再使用会方便一些&#xff0c;就是将Linux下vim打造成C IDE的样子。自动配置vim vim1.vim的基本概念2.vim的基本操作2.1进入vim编辑界面2.2 如何在vim编辑代码2.3 退出vim并保存代码2.4一张图总结基本操作3.vim正常模式命令集3.1 进入插入模式和退出插入模式…
阅读更多...
自己电脑中安装黑群辉NAS

自己电脑中安装黑群辉NAS

前期准备&#xff1a;一个品牌U盘(制作后就是启动盘需一直插电脑上)、一台安装群辉的电脑&#xff08;可以没有硬盘&#xff09;、一台可正常开机的电脑 资源下载&#xff1a; 百度网盘链接&#xff1a;链接: https://pan.baidu.com/s/1t_yVON16Pt8H1ytpvf0J-A?pwdxe7m 提取…
阅读更多...
Little Snitch 5 - Mac 老牌防火墙安全工具软件小飞贼，监控和组织特定软件的网络连接

Little Snitch 5 - Mac 老牌防火墙安全工具软件小飞贼，监控和组织特定软件的网络连接

Little Snitch 5 - Mac 老牌防火墙安全工具软件小飞贼&#xff0c;监控和组织特定软件的网络连接 一旦连接到Internet&#xff0c;应用程序就可以随时随地发送它们想要的任何东西。通常情况下&#xff0c;他们是为你做的。但有时&#xff0c;例如在跟踪软件、木马或其他恶意软件…
阅读更多...
ICLR2022 | ViT-VQGAN+：Vector-quantized Image Modeling with Improved VQGAN

ICLR2022 | ViT-VQGAN+：Vector-quantized Image Modeling with Improved VQGAN

论文链接&#xff1a;https://openreview.net/forum?idpfNyExj7z2 | https://arxiv.53yu.com/abs/2110.04627原文标题&#xff1a;Vector-quantized Image Modeling with Improved VQGAN一、问题提出Natural language processing (NLP) has recently experienced dramatic imp…
阅读更多...
程序结构你都懂了吗

程序结构你都懂了吗

上一篇&#xff1a;Python中基本输入和输出 昨天有粉丝问我&#xff0c;在看别人代码时总感觉脑子很混乱&#xff0c;不知道代码到哪步是停止&#xff0c;哪步又是开始&#xff0c;这是怎么回事呢&#xff1f; 其实很简单&#xff0c;因为还没有明白程序的基本执行流程&#x…
阅读更多...
空间滤波基础

空间滤波基础

空间滤波是一种图像处理技术&#xff0c;它通过对每个像素周围的像素进行加权平均来平滑图像。这个过程的基本思想是&#xff0c;将每个像素的灰度值与它周围像素的灰度值进行加权平均&#xff0c;然后用平均值来替换原来的像素值。空间滤波器的大小和形状决定了每个像素的加权…
阅读更多...
如何用 Redis 统计用户访问量？

如何用 Redis 统计用户访问量？

本文已经收录到Github仓库&#xff0c;该仓库包含计算机基础、Java基础、多线程、JVM、数据库、Redis、Spring、Mybatis、SpringMVC、SpringBoot、分布式、微服务、设计模式、架构等核心知识点&#xff0c;欢迎star~ Github地址&#xff1a;https://github.com/Tyson0314/Java-…
阅读更多...
DCN v1 可变形卷积v1解析（修正篇）

DCN v1 可变形卷积v1解析（修正篇）

在两年前的这篇文章Deformable Convolution&#xff08;可变形卷积&#xff09;代码解析&#xff08;有错误&#xff0c;修改中&#xff09;中&#xff0c;当时对可变形卷积进行了代码解读&#xff0c;后来被网友指出其中的解释是错的&#xff0c;里面引用的keras版本的代码实现…
阅读更多...
002、捕鱼和分鱼问题

002、捕鱼和分鱼问题

002、【题目】捕鱼和分鱼问题 捕鱼和分鱼&#xff1a;A、B、C、D、E 五个人在某天夜里合伙去捕鱼&#xff0c; 到第二天凌晨时都疲惫不堪&#xff0c;于是各自找地方睡觉。 日上三杆&#xff0c;A第一个醒来&#xff0c;他将鱼分为五份&#xff0c;把多余的一条鱼 扔掉&…
阅读更多...
C语言-数据的存储-整形的存储（8.1）

C语言-数据的存储-整形的存储（8.1）

目录 思维导图&#xff1a; 1.数据类型的基本归类 1.1类型的意义 1.2整形家族 1.3浮点数家族 1.4构造类型 1.5指针类型 1.6空类型 2. 整形在内存中的存储 2.1 原码、反码、补码 2.2 大小端介绍 2.3 练习、巩固、提高 写在最后&#xff1a; 思维导图&#xff1a; 1…
阅读更多...
视觉slam中的相机类型

视觉slam中的相机类型

作者&#xff1a;朱金灿 来源&#xff1a;clever101的专栏 为什么大多数人学不会人工智能编程&#xff1f;>>> 顾名思义&#xff0c;视觉 SLAM&#xff08;又称 vSLAM&#xff09;使用从相机和其他图像传感器采集的图像。视觉 SLAM 可以使用普通相机&#xff08;广角…
阅读更多...
【UnLua】深入理解 UnLua

【UnLua】深入理解 UnLua

【UnLua】深入理解 UnLua 从 UnLua 框架层面讨论真正值得关注的关键点 UnLua 架构UnLua 内存管理UnLua 性能 大纲 UnLua 静态导出UnLua 架构UnLua 内存管理UnLua 性能 静态导出 静态导出&#xff0c;这是标准的 Lua 用法&#xff0c;已经非常完善了&#xff0c;就一种标准…
阅读更多...
qemu virtio设备模拟与初始化流程

qemu virtio设备模拟与初始化流程

文章目录VirtIO设备模拟及初始化流程Virtio设备的创建参数解析virtio 设备初始化流程pci_bus_matchpci_match_devicepci_device_probevirtio_pci_proberegister_virtio_devicevirtio_dev_matchvirtio_dev_probe参考VirtIO设备模拟及初始化流程 qemu设备虚拟机化的路线可以概括…
阅读更多...
C++之智能指针

C++之智能指针

文章目录一、为什么需要智能指针&#xff1f;二、智能指针的使用及原理1. RAII2.智能指针的原理3. auto_ptr4. unique_ptr5. shared_ptr6. weak_ptr7.删除器一、为什么需要智能指针&#xff1f; 如果在 div() 输入的 b 0&#xff0c;那么就会抛出一个异常&#xff0c;被 main…
阅读更多...
Redis面试题总结

Redis面试题总结

一、Redis概述 1.什么是Redis&#xff1f; Redis是一个key-value存储系统&#xff0c;它支持存储的value类型相对更多&#xff0c;包括string、list、set、zset&#xff08;sorted set --有序集合&#xff09;和hash。这些数据结构都支持push/pop、add/remove及取交集并集和…
阅读更多...
[程序设计]-基于人工智能博弈树，极大极小(Minimax)搜索算法并使用Alpha-Beta剪枝算法优化实现的可人机博弈的AI智能五子棋游戏。

[程序设计]-基于人工智能博弈树，极大极小(Minimax)搜索算法并使用Alpha-Beta剪枝算法优化实现的可人机博弈的AI智能五子棋游戏。

绪论-五子棋的特点与规则 五子棋是两方之间进行的竞技活动&#xff0c;专用棋盘为15*15&#xff0c;五连子的方向为横、竖、斜&#xff1b;任一方在棋盘上形成横向、竖向、斜向的连续的相同颜色的五个&#xff08;含五个以上&#xff09;时即为该方胜利&#xff1b;在棋盘上以…
阅读更多...
Intel OneApi Developer Tools

Intel OneApi Developer Tools

“英特尔OneApi开发人员工具”是一组工具和库&#xff0c;用于为Internet发布的各种处理建筑开发高速应用程序。oneAPI是一个完全开放的编写程序模型&#xff0c;支持具有不同架构的各种制造商。使用此工具&#xff0c;其他开发人员需要为每个架构师使用特定的代码&#xff0c;…
阅读更多...
推荐文章
最新文章

Copyright @ 2022~2023