{ "cells": [ { "cell_type": "markdown", "metadata": {}, "source": [ "### 0.补充\n", "在介绍高斯混合模型的变分推断之前先补充两个分布的知识,一个是高斯-Wishart分布,它是高维高斯分布的共轭先验,另一个是学生t分布,它是高斯-Gamma分布对精度积分的边缘概率分布,下面分别介绍\n", "\n", "#### 高斯-Wishart分布\n", "\n", "对于$D$维向量$x$的多元高斯分布$N(x\\mid\\mu,\\Lambda^{-1})$,假设精度$\\Lambda$已知,则均值$\\mu$的共轭先验分布仍为高斯分布,但对于均值已知,精度未知的情况,其共轭先验是Wishart分布,定义如下: \n", "\n", "$$\n", "\\mathcal{W}(\\Lambda\\mid W,\\mathcal{v})=B|\\Lambda|^{\\frac{\\mathcal{v}-D-1}{2}}exp(-\\frac{1}{2}Tr(W^{-1}\\Lambda))\n", "$$ \n", "\n", "其中,$\\mathcal{v}$被称为分布的自由度数量,$W$是一个$D\\times D$的标量矩阵,$Tr(\\cdot)$表示矩阵的迹,归一化系数$B$为: \n", "\n", "$$\n", "B(W,\\mathcal{v})=|W|^{-\\frac{\\mathcal{v}}{2}}(2^{\\frac{\\mathcal {v}D}{2}}\\pi^{\\frac{D(D-1)}{4}}\\prod_{i=1}^D\\Gamma(\\frac{\\mathcal{v}+1-i}{2}))^{-1}\n", "$$ \n", "\n", "而对于均值和精度均未知的情况,那么类似于一元变量的推理方法,可以得到其共轭先验为: \n", "\n", "$$\n", "p(\\mu,\\Lambda\\mid\\mu_0,beta,W,\\mathcal{v})=N(\\mu\\mid\\mu_0,(\\beta\\Lambda)^{-1})\\mathcal{W}(\\Lambda\\mid W,\\mathcal{v})\n", "$$ \n", "\n", "该分布便是高斯-Wishart分布,或者正态-Wishart分布\n", "\n", "#### 学生t分布\n", "根据之前的推导我们知道,对于$\\mu$已知,$\\tau$未知的一元高斯分布$N(x\\mid\\mu,\\tau^{-1})$,它的共轭分布为Gamma分布,不妨表示为$Gam(\\tau\\mid a,b)$,如果把精度积分掉,那么就得到了一个边缘概率分布: \n", "\n", "$$\n", "p(x\\mid \\mu,a,b)=\\int_0^\\infty N(x\\mid\\mu,\\tau^{-1})Gam(\\tau\\mid a,b)d\\tau\\\\\n", "=\\int_0^\\infty \\frac{b^ae^{(-b\\tau)}\\tau^{a-1}}{\\Gamma(a)}(\\frac{\\tau}{2\\pi})^{\\frac{1}{2}}exp(-\\frac{\\tau}{2}(x-\\mu)^2)d\\tau\\\\\n", "=\\frac{b^a}{\\Gamma(a)}(\\frac{1}{2\\pi})^{\\frac{1}{2}}\\int_0^\\infty(\\frac{z}{b+\\frac{(x-\\mu)^2}{2}})^{a-\\frac{1}{2}}e^{-z}[b+\\frac{(x-\\mu)^2}{2}]^{-1}dz(令z=\\tau(b+\\frac{(x-\\mu)^2}{2}))\\\\\n", "=\\frac{b^a}{\\Gamma(a)}(\\frac{1}{2\\pi})^{\\frac{1}{2}}(b+\\frac{(x-\\mu)^2}{2})^{-a-\\frac{1}{2}}\\int_0^\\infty z^{(a+\\frac{1}{2})-1}e^{-z}dz\\\\\n", "=\\frac{b^a}{\\Gamma(a)}(\\frac{1}{2\\pi})^{\\frac{1}{2}}(b+\\frac{(x-\\mu)^2}{2})^{-a-\\frac{1}{2}}\\Gamma(a+\\frac{1}{2})\n", "$$ \n", "\n", "我们对参数做一下调整,令$\\mathcal{v}=2a,\\lambda=\\frac{a}{b}$,那么这就是学生t分布: \n", "\n", "$$\n", "St(x\\mid\\mu,\\lambda,\\mathcal{v})=\\frac{\\Gamma(\\frac{\\mathcal{v}}{2}+\\frac{1}{2})}{\\Gamma(\\frac{\\mathcal{v}}{2})}(\\frac{\\lambda}{\\pi\\mathcal{v}})^{\\frac{1}{2}}(1+\\frac{\\lambda(x-\\mu)^2}{\\mathcal{v}})^{-\\frac{\\mathcal{v}}{2}-\\frac{1}{2}}\n", "$$ \n", "\n", "这里,参数$\\lambda$被称为t分布的精度(注意:通常情况下,它不是方差的倒数),参数$\\mathcal{v}$被称为自由度,它的作用如下图所示(其中$\\mu=0,\\lambda=1$),对于$\\mathcal{v}=1$的情况下,t分布是柯西分布,而对于$\\mathcal{v}\\rightarrow \\infty$的情况下,t分布$St(x\\mid\\mu,\\lambda,\\mathcal{v})$就变成了高斯分布$N(x\\mid\\mu,\\lambda^{-1})$ \n", "\n", "\n", "![avatar](./source/15_t分布1.png)\n", "\n", "需要注意的一点是,学生t分布可以看作无限个同均值不同精度的高斯分布相加所得到(高斯混合模型的极限),所以t分布通常有更长的“尾巴”,建模时具有更强的鲁棒性,如下图是用t分布(红线)和高斯分布(绿线)建模的对比,在未引入噪声数据之前,如(a)图,t分布和高斯分布的建模效果相同,而在引入噪声数据之后,如(b)图,会对高斯分布模型造成较大的影响,而t分布几乎不受影响\n", "\n", "![avatar](./source/15_t分布2.png) \n", "\n", "\n", "高维学生t分布: \n", "\n", "将一维学生t分布扩展到高维,可以写作 \n", "\n", "$$\n", "St(x\\mid\\mu,\\Lambda,\\mathcal{v})=\\frac{\\Gamma(\\frac{D}{2}+\\frac{\\mathcal{v}}{2})}{\\Gamma(\\frac{\\mathcal{v}}{2})}\\frac{|\\Lambda|^{\\frac{1}{2}}}{(\\pi\\mathcal{v})^{\\frac{D}{2}}}[1+\\frac{(x-\\mu)^T\\Lambda(x-\\mu)}{\\mathcal{v}}]^{-\\frac{D}{2}-\\frac{\\mathcal{v}}{2}}\n", "$$\n", "\n", "其中,$D$是$x$的维度,另外t分布具有如下的性质: \n", "\n", "(1)$E[x]=\\mu,\\mathcal{v}>1$; \n", "\n", "(2)$cov[x]=\\frac{\\mathcal{v}}{\\mathcal{v}-2}\\Lambda^{-1},\\mathcal{v}>2$; \n", "\n", "(3)$mode[x]=\\mu$ \n", "\n", "好的,补充内容差不多就是这些了,接下来按照上一节总结的四个步骤来介绍高斯混合模型的变分推断" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### 一.联合概率分布\n", "高斯混合模型的图模型如下图\n", "![avatar](./source/15_VI_GMM1.png)\n", "\n", "这里是概率图的盘式记法,其中,紫色实心的圈表示观测变量,红色空心圈表示隐变量(参数),框(盘子)内表示$N$次独立同分布的观测,接下来首先从似然函数出发,对于我们的每个观测$x_n$,都有一个对应的隐变量$z_n$,它是一个$K$维的二值向量,每个元素为$z_{nk},k=1,2,...,K$,表示对应的观测$x_n$是否采样于第$k$个高斯分布,对于给定的混合系数$\\pi=\\{\\pi_1,\\pi_2,...,\\pi_K\\}$(即每个高斯分布的权重系数),可以写出$Z={z_1,z_2,...,z_N}$关于$\\pi$的条件概率分布: \n", "\n", "$$\n", "p(Z\\mid\\pi)=\\prod_{n=1}^N\\prod_{k=1}^K\\pi_k^{z_{nk}}\n", "$$ \n", "\n", "同样地,在给定了$Z,\\mu,\\Lambda=\\Sigma^{-1}$的条件下,我们可以写出观测数据的条件概率分布: \n", "\n", "$$\n", "p(X\\mid Z,\\mu,\\Lambda)=\\prod_{n=1}^N\\prod_{k=1}^KN(x_n\\mid\\mu_k,\\Lambda_k^{-1})^{z_{nk}}\n", "$$ \n", "\n", "这里,$\\mu=\\{\\mu_k\\},\\Lambda=\\{\\Lambda_k\\}$,这里使用精度矩阵是为了后面的计算方便,接下来引入参数$\\mu,\\Lambda,\\pi$上的先验概率分布,显然为了计算简单,我们可以使用共轭先验,对于混合系数$\\pi$的先验概率,我们可以使用狄利克雷分布: \n", "\n", "$$\n", "p(\\pi)=Dir(\\pi\\mid\\alpha_0)=C(\\alpha_0)\\prod_{k=1}^K\\pi_k^{\\alpha_0-1}\n", "$$ \n", "\n", "这里,我们假设每个高斯分布的被观测的先验数量都一样为$\\alpha_0$,$C(\\alpha_0)$表示归一化系数,而对于$\\mu,\\Lambda$可以选择高斯-Wishart先验分布: \n", "\n", "$$\n", "p(\\mu,\\Lambda)=p(\\mu\\mid\\Lambda)p(\\Lambda)\\\\\n", "=\\prod_{k=1}^KN(\\mu_k\\mid m_0,(\\beta_0\\Lambda_k)^{-1})\\mathcal{W}(\\Lambda_k\\mid W_0,\\mathcal{v}_0)\n", "$$ \n", "\n", "为了对称性,通常取$m_0=0$,注意,由于高斯-Wishart分布中$\\mu$的概率分布的方差是$\\Lambda$的函数,所以,这时的概率图模型需要被更新为如下形式: \n", "![avatar](./source/15_VI_GMM2.png) \n", "\n", "根据概率图模型,我们将上面的几项乘起来就是联合概率分布咯: \n", "\n", "$$\n", "p(X,Z,\\pi,\\mu,\\Lambda)=p(X\\mid Z,\\mu,\\Lambda)p(Z\\mid\\pi)p(\\pi)p(\\mu\\mid\\Lambda)p(\\Lambda)\n", "$$" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### 二.变分分布\n", "上面推导出了联合概率分布,那么接下来就需要造一个变分分布去近似它的后验概率分布咯: \n", "\n", "$$\n", "q(Z,\\pi,\\mu,\\Lambda)\\rightarrow p(Z,\\pi,\\mu,\\Lambda\\mid X)\n", "$$ \n", "\n", "接下来,我们做一个假设:变分分布可以在隐变量和参数之间记性分解,即: \n", "\n", "$$\n", "q(Z,\\pi,\\mu,\\Lambda)=q(Z)q(\\pi,\\mu,\\Lambda)\n", "$$ \n", "\n", "注意,这里省略了下标,$q(Z)$与$q(\\pi,\\mu,\\Lambda)$是不一样的分布,我们通过参数来区分不同的分布,接下来,就有: \n", "\n", "$$\n", "ln\\ q^*(Z)=E_{\\pi,\\mu,\\Lambda}[ln\\ p(X,Z,\\pi,\\mu,\\Lambda)]+const\n", "$$ \n", "\n", "将上面,联合概率分布的表达式带入可得(将与$Z$无关的项整合到$const$中): \n", "\n", "$$\n", "ln\\ q^*(Z)=E_\\pi[ln\\ p(Z\\mid\\pi)]+E_{\\mu,\\Lambda}[ln\\ p(X\\mid Z,\\mu,\\Lambda)]+const\n", "$$ \n", "\n", "替换右侧的两个条件分布,然后再次把与$Z$无关的项整合到$const$中: \n", "\n", "$$\n", "ln\\ q^*(Z)=\\sum_{n=1}^N\\sum_{k=1}^Kz_{nk}ln\\ \\rho_{nk}+const\n", "$$ \n", "\n", "其中: \n", "$$\n", "ln\\ \\rho_{nk}=E[ln\\ \\pi_k]+\\frac{1}{2}E[ln\\ |\\Lambda_k|]-\\frac{D}{2}ln(2\\pi)-\\frac{1}{2}E_{\\mu_k,\\Lambda_k}[(x_n-\\mu_k)^T\\Lambda_k(x_n-\\mu_k)]\n", "$$ \n", "\n", "其中,$D$是观测变量$x$的维度,对上面的等式两边取指数,有: \n", "\n", "$$\n", "q^*(Z)\\propto\\prod_{n=1}^N\\prod_{k=1}^K\\rho_{nk}^{z_{nk}}\n", "$$ \n", "\n", "由于,$q^*(Z)$需要是归一化的,而且同时$z_{nk},k=1,2,...,K$是二值的($K$个数中只有一个为1,其余为0),所以,我们有: \n", "\n", "$$\n", "q^*(Z)=\\prod_{n=1}^N\\prod_{k=1}^Kr_{nk}^{z_{nk}}\\\\\n", "r_{nk}=\\frac{\\rho_{nk}}{\\sum_{j=1}^K\\rho_{nj}}\n", "$$ \n", "\n", "对于离散情况有: \n", "\n", "$$\n", "E[z_{nk}]=r_{nk}\n", "$$ \n", "\n", "可以发现,$q^*(Z)$的最优解是依赖于其他变量的,即它们是耦合的,需要使用迭代的方式进行求解,为了后面的推导方便,我们定义如下的三个量: \n", "\n", "$$\n", "N_k=\\sum_{n=1}^Nr_{nk}\\\\\n", "\\bar{x}_k=\\frac{1}{N_k}\\sum_{n=1}^Nr_{nk}x_n\\\\\n", "S_k=\\frac{1}{N_k}\\sum_{n=1}^Nr_{nk}(x_n-\\bar{x}_k)(x_n-\\bar{x}_k)^T\n", "$$ \n", "\n", "大家可以发现,这里定义的这些量其实和EM算法中计算用到的量类似,接下来看看$q(\\pi,\\mu,\\Lambda)$: \n", "\n", "$$\n", "ln\\ q^*(\\pi,\\mu,\\Lambda)=ln\\ p(\\pi)+\\sum_{k=1}^Kln\\ p(\\mu_k,\\Lambda_k)+E_Z[ln\\ p(Z\\mid\\pi)]+\\sum_{k=1}^K\\sum_{n=1}^NE[z_{nk}]ln\\ N(x_n\\mid\\mu_k,\\Lambda_k^{-1})+const\n", "$$ \n", "\n", "可以发现,表达式右边中一些项只与$\\pi$相关,另外一些项只与$\\mu$和$\\Lambda$相关,也就是说$q(\\pi,\\mu,\\Lambda)$还可以进一步分解为$q(\\pi)q(\\mu,\\Lambda)$,而$q(\\mu,\\Lambda)$由可以分解为$k$与$\\mu_k,\\Lambda_k$相关的项,所以: \n", "\n", "$$\n", "q(\\pi,\\mu,\\Lambda)=q(\\pi)\\prod_{k=1}^Kq(\\mu_k,\\Lambda_k)\n", "$$ \n", "\n", "分离出与$\\pi$相关的项,我们有: \n", "\n", "$$\n", "ln\\ q^*(\\pi)=(\\alpha_0-1)\\sum_{k=1}^Kln\\ \\pi_k+\\sum_{k=1}^K\\sum_{n=1}^Nr_{nk}ln\\ \\pi_k+const\n", "$$ \n", "\n", "对上面的表达式两侧取指数可以发现,$q^*(\\pi)$是一个狄利克雷分布: \n", "\n", "$$\n", "q^*(\\pi)=Dir(\\pi\\mid\\alpha),\\alpha_k=\\alpha_0+N_k\n", "$$ \n", "\n", "最后,我们可以推出$q^*(\\mu_k,\\Lambda_k)$是一个高斯-Wishart分布: \n", "\n", "$$\n", "q^*(\\mu_k,\\Lambda_k)=N(\\mu_k\\mid m_k,(\\beta_k\\Lambda_k)^{-1})\\mathcal{W}(\\Lambda_k\\mid W_k,\\mathcal{v}_k)\n", "$$ \n", "\n", "其中: \n", "\n", "$$\n", "\\beta_k=\\beta_0+N_k\\\\\n", "m_k=\\frac{1}{\\beta_k}(\\beta_0m_0+N_k\\bar{x}_k)\\\\\n", "W_k^{-1}=W_0^{-1}+N_kS_k+\\frac{\\beta_0N_k}{\\beta_0+N_k}(\\bar{x}_k-m_0)(\\bar{x}_k-m_0)^T\\\\\n", "\\mathcal{v}_k=\\mathcal{v}_0+N_k\n", "$$" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### 三.迭代更新\n", "\n", "根据上面的推导,迭代更新流程就知道了:$r_{nk}\\rightarrow \\pi_k,\\mu_k,\\Lambda_k\\rightarrow r_{nk}\\rightarrow \\cdots$,剩余还需要进一步求解的便是$r_{nk}$被归一化前的变量$\\rho_{nk}$,它的计算公式,我们再写一下: \n", "\n", "$$\n", "ln\\ \\rho_{nk}=E[ln\\ \\pi_k]+\\frac{1}{2}E[ln\\ |\\Lambda_k|]-\\frac{D}{2}ln(2\\pi)-\\frac{1}{2}E_{\\mu_k,\\Lambda_k}[(x_n-\\mu_k)^T\\Lambda_k(x_n-\\mu_k)]\n", "$$ \n", "\n", "根据Wishart分布,狄利克雷分布的特性,我们可以得到: \n", "\n", "$$\n", "E[ln\\ |\\Lambda_k|]=\\sum_{i=1}^D\\psi(\\frac{\\mathcal{v}_k+1-i}{2})+Dln2+ln|W_k|\\\\\n", "E[ln\\ \\pi_k]=\\psi(\\alpha_k)-\\psi(\\hat{\\alpha})\n", "$$ \n", "\n", "这里$\\hat{\\alpha}=\\sum_k\\alpha_k$,$\\psi(\\cdot)$表示Digamma函数: \n", "$$\n", "\\psi(a)=\\frac{d}{da}ln\\Gamma(a)\n", "$$ \n", "\n", "另外,由高斯-Wishart分布,可以求得: \n", "\n", "\n", "$$\n", "E_{\\mu_k,\\Lambda_k}[(x_n-\\mu_k)^T\\Lambda_k(x_n-\\mu_k)]=D\\beta_k^{-1}+\\mathcal{v}_k(x_n-m_k)^TW_k(x_n-m_k)\n", "$$\n" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### 四.变分分布的应用:概率密度预测\n", "对于GMM,我们通常对新样本$\\hat{x}$的概率密度感兴趣,根据贝叶斯推断理论,它的预测过程可以写作: \n", "\n", "$$\n", "p(\\hat{x}\\mid X)=\\sum_{\\hat{z}}\\int\\int\\int p(\\hat{x}\\mid \\hat{z},\\mu,\\Lambda)p(\\hat{z}\\mid \\pi)p(\\pi,\\mu,\\Lambda\\mid X)d\\pi d\\mu d\\Lambda\\\\\n", "=\\sum_{k=1}^K\\int\\int\\int \\pi_kN(\\hat{x}\\mid \\mu_k,\\Lambda_k^{-1})p(\\pi,\\mu,\\Lambda\\mid X)d\\pi d\\mu d\\Lambda(消去\\hat{x})\\\\\n", "\\simeq \\sum_{k=1}^K\\int\\int\\int \\pi_kN(\\hat{x}\\mid \\mu_k,\\Lambda_k^{-1})q(\\pi)q(\\mu_k,\\Lambda_k)d\\pi d\\mu_k d\\Lambda_k(变分分布替换后验分布)\\\\\n", "=\\frac{1}{\\hat{\\alpha}}\\sum_{k=1}^K\\alpha_kSt(\\hat{x}\\mid m_k,L_k,\\mathcal{v}_k+1-D)\n", "$$ \n", "\n", "我们最终发现,概率密度其实是学生t分布的混合,其中: \n", "\n", "$$\n", "L_k=\\frac{(\\mathcal{v}_k+1-D)\\beta_k}{1+\\beta_k}W_k\n", "$$ \n", "\n", "各符号的意义与前面推导的一致" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### 五.代码实现\n", "首先,我们将几个分布的函数写一下" ] }, { "cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [], "source": [ "import numpy as np\n", "from scipy import special\n", "import matplotlib.pyplot as plt\n", "%matplotlib inline" ] }, { "cell_type": "code", "execution_count": 2, "metadata": {}, "outputs": [], "source": [ "\"\"\"\n", "代码封装到utils中\n", "\"\"\"\n", "\n", "def dirichlet(u, alpha):\n", " \"\"\"\n", " 狄利克雷分布\n", " :param u: 随机变量\n", " :param alpha: 超参数\n", " :return:\n", " \"\"\"\n", " # 计算归一化因子\n", " beta = special.gamma(np.sum(alpha))\n", " for alp in alpha:\n", " beta /= special.gamma(np.sum(alp))\n", " rst = beta\n", " # 计算结果\n", " for idx in range(0, len(alpha)):\n", " rst *= np.power(u[idx], alpha[idx] - 1)\n", " return rst\n", "\n", "\n", "def wishart(Lambda, W, v):\n", " \"\"\"\n", " wishart分布\n", " :param Lambda:随机变量\n", " :param W:超参数\n", " :param v:超参数\n", " :return:\n", " \"\"\"\n", " # 维度\n", " D = W.shape[0]\n", " # 先计算归一化因子\n", " B = np.power(np.linalg.det(W), -1 * v / 2)\n", " B_ = np.power(2.0, v * D / 2) * np.power(np.pi, D * (D - 1) / 4)\n", " for i in range(1, D + 1):\n", " B_ *= special.gamma((v + 1 - i) / 2)\n", " B = B / B_\n", " # 计算剩余部分\n", " rst = B * np.power(np.linalg.det(Lambda), (v - D - 1) / 2)\n", " rst *= np.exp(-0.5 * np.trace(np.linalg.inv(W) @ Lambda))\n", " return rst\n", "\n", "\n", "def St(X, mu, Lambda, v):\n", " \"\"\"\n", " 学生t分布\n", " :param X: 随机变量\n", " :param mu: 超参数\n", " :param Lambda: 超参数\n", " :param v: 超参数\n", " :return:\n", " \"\"\"\n", " n_sample, D = X.shape\n", " return special.gamma(D / 2 + v / 2) * np.power(np.linalg.det(Lambda), 0.5) * np.power(\n", " 1 + np.sum((X - mu) @ Lambda * (X - mu), axis=1) / v, -1.0 * D / 2 - v / 2) / special.gamma(v / 2) / np.power(\n", " np.pi * v, D / 2)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "接下里就是模型代码了..." ] }, { "cell_type": "code", "execution_count": 3, "metadata": {}, "outputs": [], "source": [ "import os\n", "os.chdir('../')\n", "from ml_models import utils\n", "\n", "\n", "\"\"\"\n", "高斯混合模型的变分推断实现,封装到ml_models.vi中\n", "\"\"\"\n", "\n", "\n", "class GMMCluster(object):\n", " def __init__(self, n_components=1, tol=1e-6, n_iter=100, prior_dirichlet_alpha=None, prior_gaussian_mean=None,\n", " prior_gaussian_beta=None, prior_wishart_w=None, prior_wishart_v=None,\n", " verbose=False):\n", " \"\"\"\n", " GMM的VI实现\n", " :param n_components: 高斯混合模型数量\n", " :param tol: -log likehold增益" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "#查看效果\n", "utils.plot_contourf(X,gmm.predict_sample_generate_proba,lines=5)" ] }, { "cell_type": "code", "execution_count": 7, "metadata": {}, "outputs": [ { "data": { "image/png": "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\n", "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "#查看分类边界\n", "utils.plot_decision_function(X,gmm.predict(X),gmm)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "可以发现VI的收敛效果还是不错的,接下来再做一个测试,假如初始的n_components设置很大会怎样" ] }, { "cell_type": "code", "execution_count": 8, "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "D:\\app\\Anaconda3\\lib\\site-packages\\matplotlib\\contour.py:967: UserWarning: The following kwargs were not used by contour: 'linewidth'\n", " s)\n" ] }, { "data": { "image/png": "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\n", "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "gmm = GMMCluster(verbose=False, n_iter=100, n_components=20)\n", "gmm.fit(X)\n", "utils.plot_contourf(X,gmm.predict_sample_generate_proba,lines=5)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "可以发现,上面设置初始聚类中心为20和设置为4的训练结果一样,说明**VI训练的GMM对超参数并不敏感**(这主要是由于引入了先验分布的缘故),所以我们不必再对超参数做交叉验证进行选择了...." ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### 六.总结一下\n", "这一节完整的使用了四个步骤来做GMM的训练,虽然结果还OK,但我们需要更加清晰的认识到一个事实:**如何对隐变量进行拆分,以及假设各变量子集服从何种分布是成功的关键**,如果拆分方式不对,或者分布假设不对可能会取得比较差的结果(并没有做实验去检验...《机器学习》书上的观点)" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [] } ], "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": "3.6.4" } }, "nbformat": 4, "nbformat_minor": 2 }