不论高手不高手，如果能够把这个东西看懂不想得分都很难

myadvice 2003-04-10 11:24:34

刚从网上认识一个大家的女孩，她告诉我她们的一道平时作业，反正我的C语言已经忘得差不多了，现在用的是Delphi，所以如果哪位想看看现在的大学怎么样看看这道题吧！按要求或者是调试正确者肯定给分，我这个人没有不结贴的习惯。

其题目要求如下：（现在学校可能升级了，连题目全都用英文了）

Modify take_e.c, the exhaustive search version of the game take, to use alpha-beta pruning and the optional argument that specifies a maximum lookahead depth.

Provide detailed inline documentation of all modifications, starting all your comments with /*! and ending them with !*/.

You will be evaluated on the

* correctness of your implementation of alpha-beta pruning

* correctness of your implementation of maximum lookahead

* quality of your functions for estimating the value of a non-terminal node

* understanding of the task demonstrated by your inline documentation

The submitted code must compile under Turbo C.

程序代码在下边，因为csdn不准贴子太长

...全文

36 25 打赏收藏转发到动态举报

写回复

用AI写文章

25 条回复

切换为时间正序

请发表友善的回复…

发表回复

lymgf 2003-04-12

打赏
举报

/* if it is the best so far, remember it and its value */
if (depth==1)
{
if (this_score > alpha) move = move_codes[move_index];
if (this_score > alpha) alpha = this_score;
}
这一段忘了去掉

lymgf 2003-04-12

打赏
举报

楼上说的不对，被评估的节点并非同时展开的，因此内存消耗只跟搜索深度相关，为线性增长。

下面是我写的alpha-beta剪枝搜索，由于不想改原程序定义的数据结构，所以写的时候总觉得别扭，另外附上do_a_move的测试函数。可以看出alpha-beta剪枝有效减少了评估的节点，但是到7*7的棋盘时，显得力不从心。如果增加其他搜索手段，如历史启发、备忘录等，可能可以对7*7搜索到根，也就不用估值函数。若棋盘继续加大，就必须限制搜索深度，这时估值函数显得极为重要,个人认为这才是最难点,就不费心写了。

/*
**alpha-beta pruning 返回评估值
*/
char move; //增加一个全局变量，表示最佳解

int alpha_beta(position_type *pos,int alpha,int beta,int depth)
{
int this_score; /* the score of the move being considered */
int move_index; /* the move to check */
position_type posn = *pos; /* copy of the current position - may be modified */

no_of_nodes++;
if (finished(&posn))
{
return ((posn.player=='x')?1:-1)*posn.score;
}

/* check each of the possible moves in turn */
for (move_index = 0; move_index < NO_OF_MOVE_CODES; move_index++)
{
/* check if the move is valid */
if (valid_move(pos, move_codes[move_index]))
{
update_position(&posn, move_codes[move_index]);
/* if so, evaluate it */
this_score = -alpha_beta(&posn, -beta,-alpha,depth+1);
/* copy of the current position - may be modified */
posn = *pos;

/* if it is the best so far, remember it and its value */
if (depth==1)
{
if (this_score > alpha) move = move_codes[move_index];
if (this_score > alpha) alpha = this_score;
}
if (this_score > alpha)
{
if (depth==1) move = move_codes[move_index];
alpha = this_score;
}
if (alpha>=beta) break;
}
}

return alpha;
}

/* do a move
** updates the global data structures representing the current position
*/
void do_a_move(void)
{
int val;

no_of_nodes=0;
move = select_best_move(&position, &val);
printf("I predict a final score of %d\n", val);
printf("Consider %ld nodes.\n",no_of_nodes);
printf("E-S suggest move:%c\n",move);

no_of_nodes=0;
val=alpha_beta(&position,-20000, 20000,1);
printf("I predict a final score of %d\n", val);
printf("Consider %ld nodes.\n",no_of_nodes);
printf("A-B suggest move:%c\n",move);

update_position(&position, move);
}

myadvice 2003-04-12