algorithm/hello-algo
1
0
Fork 0
mirror of https://github.com/krahets/hello-algo.git synced 2025-07-24 18:55:36 +08:00
Code Issues Packages Projects Releases Wiki Activity

deploy

Browse Source
This commit is contained in:
krahets
2023-09-22 13:08:10 +08:00
parent 5bb9f76fbc
commit 6fffa33695
107 changed files with 2561 additions and 19178 deletions
Download Patch File Download Diff File Expand all files Collapse all files

216
chapter_dynamic_programming/intro_to_dynamic_programming/index.html
View File

@ -60,7 +60,18 @@
</head>
<link href="../../assets/stylesheets/glightbox.min.css" rel="stylesheet"/><style>
html.glightbox-open { overflow: initial; height: 100%; }
.gslide-title { margin-top: 0px; user-select: text; }
.gslide-desc { color: #666; user-select: text; }
.gslide-image img { background: white; }
.gscrollbar-fixer { padding-right: 15px; }
.gdesc-inner { font-size: 0.75rem; }
body[data-md-color-scheme="slate"] .gdesc-inner { background: var(--md-default-bg-color);}
body[data-md-color-scheme="slate"] .gslide-title { color: var(--md-default-fg-color);}
body[data-md-color-scheme="slate"] .gslide-desc { color: var(--md-default-fg-color);}
</style> <script src="../../assets/javascripts/glightbox.min.js"></script></head>
@ -1875,14 +1886,6 @@
10.2 &nbsp; 二分查找插入点
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
</li>
@ -1903,14 +1906,6 @@
10.3 &nbsp; 二分查找边界
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
</li>
@ -2330,14 +2325,6 @@
第 12 章 &nbsp; 分治
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
@ -2369,14 +2356,6 @@
12.1 &nbsp; 分治算法
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
</li>
@ -2397,14 +2376,6 @@
12.2 &nbsp; 分治搜索策略
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
</li>
@ -2425,14 +2396,6 @@
12.3 &nbsp; 构建树问题
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
</li>
@ -2453,14 +2416,6 @@
12.4 &nbsp; 汉诺塔问题
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
</li>
@ -2481,14 +2436,6 @@
12.5 &nbsp; 小结
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
</li>
@ -2722,14 +2669,6 @@
第 14 章 &nbsp; 动态规划
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
@ -2770,14 +2709,6 @@
14.1 &nbsp; 初探动态规划
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
<span class="md-nav__icon md-icon"></span>
</label>
@ -2789,14 +2720,6 @@
14.1 &nbsp; 初探动态规划
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
@ -2865,14 +2788,6 @@
14.2 &nbsp; DP 问题特性
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
</li>
@ -2893,14 +2808,6 @@
14.3 &nbsp; DP 解题思路
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
</li>
@ -2921,14 +2828,6 @@
14.4 &nbsp; 0-1 背包问题
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
</li>
@ -2949,14 +2848,6 @@
14.5 &nbsp; 完全背包问题
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
</li>
@ -2977,14 +2868,6 @@
14.6 &nbsp; 编辑距离问题
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
</li>
@ -3005,14 +2888,6 @@
14.7 &nbsp; 小结
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
</li>
@ -3071,14 +2946,6 @@
第 15 章 &nbsp; 贪心
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
@ -3110,14 +2977,6 @@
15.1 &nbsp; 贪心算法
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
</li>
@ -3138,14 +2997,6 @@
15.2 &nbsp; 分数背包问题
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
</li>
@ -3166,14 +3017,6 @@
15.3 &nbsp; 最大容量问题
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
</li>
@ -3194,14 +3037,6 @@
15.4 &nbsp; 最大切分乘积问题
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
</li>
@ -3222,14 +3057,6 @@
15.5 &nbsp; 小结
</span>
<span class="md-status md-status--new" title="最近添加">
</span>
</a>
</li>
@ -3517,7 +3344,7 @@
<p>给定一个共有 <span class="arithmatex">\(n\)</span> 阶的楼梯，你每步可以上 <span class="arithmatex">\(1\)</span> 阶或者 <span class="arithmatex">\(2\)</span> 阶，请问有多少种方案可以爬到楼顶。</p>
</div>
<p>如图 14-1 所示，对于一个 <span class="arithmatex">\(3\)</span> 阶楼梯，共有 <span class="arithmatex">\(3\)</span> 种方案可以爬到楼顶。</p>
<p><img alt="爬到第 3 阶的方案数量" src="../intro_to_dynamic_programming.assets/climbing_stairs_example.png" /></p>
<p><a class="glightbox" href="../intro_to_dynamic_programming.assets/climbing_stairs_example.png" data-type="image" data-width="100%" data-height="auto" data-desc-position="bottom"><img alt="爬到第 3 阶的方案数量" src="../intro_to_dynamic_programming.assets/climbing_stairs_example.png" /></a></p>
<p align="center"> 图 14-1 &nbsp; 爬到第 3 阶的方案数量 </p>
<p>本题的目标是求解方案数量，<strong>我们可以考虑通过回溯来穷举所有可能性</strong>。具体来说，将爬楼梯想象为一个多轮选择的过程：从地面出发，每轮选择上 <span class="arithmatex">\(1\)</span> 阶或 <span class="arithmatex">\(2\)</span> 阶，每当到达楼梯顶部时就将方案数量加 <span class="arithmatex">\(1\)</span> ，当越过楼梯顶部时就将其剪枝。</p>
@ -3852,7 +3679,7 @@ dp[i-1], dp[i-2], \dots, dp[2], dp[1]
dp[i] = dp[i-1] + dp[i-2]
\]</div>
<p>这意味着在爬楼梯问题中，各个子问题之间存在递推关系，<strong>原问题的解可以由子问题的解构建得来</strong>。图 14-2 展示了该递推关系。</p>
<p><img alt="方案数量递推关系" src="../intro_to_dynamic_programming.assets/climbing_stairs_state_transfer.png" /></p>
<p><a class="glightbox" href="../intro_to_dynamic_programming.assets/climbing_stairs_state_transfer.png" data-type="image" data-width="100%" data-height="auto" data-desc-position="bottom"><img alt="方案数量递推关系" src="../intro_to_dynamic_programming.assets/climbing_stairs_state_transfer.png" /></a></p>
<p align="center"> 图 14-2 &nbsp; 方案数量递推关系 </p>
<p>我们可以根据递推公式得到暴力搜索解法。以 <span class="arithmatex">\(dp[n]\)</span> 为起始点，<strong>递归地将一个较大问题拆解为两个较小问题的和</strong>，直至到达最小子问题 <span class="arithmatex">\(dp[1]\)</span> 和 <span class="arithmatex">\(dp[2]\)</span> 时返回。其中，最小子问题的解是已知的，即 <span class="arithmatex">\(dp[1] = 1\)</span>、<span class="arithmatex">\(dp[2] = 2\)</span> ，表示爬到第 <span class="arithmatex">\(1\)</span>、<span class="arithmatex">\(2\)</span> 阶分别有 <span class="arithmatex">\(1\)</span>、<span class="arithmatex">\(2\)</span> 种方案。</p>
@ -4052,7 +3879,7 @@ dp[i] = dp[i-1] + dp[i-2]
</div>
</div>
<p>图 14-3 展示了暴力搜索形成的递归树。对于问题 <span class="arithmatex">\(dp[n]\)</span> ，其递归树的深度为 <span class="arithmatex">\(n\)</span> ，时间复杂度为 <span class="arithmatex">\(O(2^n)\)</span> 。指数阶属于爆炸式增长，如果我们输入一个比较大的 <span class="arithmatex">\(n\)</span> ，则会陷入漫长的等待之中。</p>
<p><img alt="爬楼梯对应递归树" src="../intro_to_dynamic_programming.assets/climbing_stairs_dfs_tree.png" /></p>
<p><a class="glightbox" href="../intro_to_dynamic_programming.assets/climbing_stairs_dfs_tree.png" data-type="image" data-width="100%" data-height="auto" data-desc-position="bottom"><img alt="爬楼梯对应递归树" src="../intro_to_dynamic_programming.assets/climbing_stairs_dfs_tree.png" /></a></p>
<p align="center"> 图 14-3 &nbsp; 爬楼梯对应递归树 </p>
<p>观察图 14-3 ，<strong>指数阶的时间复杂度是由于“重叠子问题”导致的</strong>。例如 <span class="arithmatex">\(dp[9]\)</span> 被分解为 <span class="arithmatex">\(dp[8]\)</span> 和 <span class="arithmatex">\(dp[7]\)</span> ，<span class="arithmatex">\(dp[8]\)</span> 被分解为 <span class="arithmatex">\(dp[7]\)</span> 和 <span class="arithmatex">\(dp[6]\)</span> ，两者都包含子问题 <span class="arithmatex">\(dp[7]\)</span> 。</p>
@ -4339,7 +4166,7 @@ dp[i] = dp[i-1] + dp[i-2]
</div>
</div>
<p>观察图 14-4 ，<strong>经过记忆化处理后，所有重叠子问题都只需被计算一次，时间复杂度被优化至 <span class="arithmatex">\(O(n)\)</span></strong> ，这是一个巨大的飞跃。</p>
<p><img alt="记忆化搜索对应递归树" src="../intro_to_dynamic_programming.assets/climbing_stairs_dfs_memo_tree.png" /></p>
<p><a class="glightbox" href="../intro_to_dynamic_programming.assets/climbing_stairs_dfs_memo_tree.png" data-type="image" data-width="100%" data-height="auto" data-desc-position="bottom"><img alt="记忆化搜索对应递归树" src="../intro_to_dynamic_programming.assets/climbing_stairs_dfs_memo_tree.png" /></a></p>
<p align="center"> 图 14-4 &nbsp; 记忆化搜索对应递归树 </p>
<h2 id="1413">14.1.3 &nbsp; 方法三：动态规划<a class="headerlink" href="#1413" title="Permanent link">&para;</a></h2>
@ -4551,7 +4378,7 @@ dp[i] = dp[i-1] + dp[i-2]
</div>
</div>
<p>图 14-5 模拟了以上代码的执行过程。</p>
<p><img alt="爬楼梯的动态规划过程" src="../intro_to_dynamic_programming.assets/climbing_stairs_dp.png" /></p>
<p><a class="glightbox" href="../intro_to_dynamic_programming.assets/climbing_stairs_dp.png" data-type="image" data-width="100%" data-height="auto" data-desc-position="bottom"><img alt="爬楼梯的动态规划过程" src="../intro_to_dynamic_programming.assets/climbing_stairs_dp.png" /></a></p>
<p align="center"> 图 14-5 &nbsp; 爬楼梯的动态规划过程 </p>
<p>与回溯算法一样，动态规划也使用“状态”概念来表示问题求解的某个特定阶段，每个状态都对应一个子问题以及相应的局部最优解。例如，爬楼梯问题的状态定义为当前所在楼梯阶数 <span class="arithmatex">\(i\)</span> 。</p>
@ -4897,10 +4724,15 @@ aria-label="页脚"
<div class="md-copyright">
<div class="md-copyright__highlight">
Copyright &copy; 2023 Krahets
Copyright &copy; 2022 - 2023 Krahets
</div>
Made with
<a href="https://squidfunk.github.io/mkdocs-material/" target="_blank" rel="noopener">
Material for MkDocs
</a>
</div>
<!-- Social links -->
@ -4969,5 +4801,5 @@ aria-label="页脚"
</body>
<script>document$.subscribe(() => {const lightbox = GLightbox({"touchNavigation": true, "loop": false, "zoomable": true, "draggable": false, "openEffect": "zoom", "closeEffect": "zoom", "slideEffect": "none"});})</script></body>
</html>