Expectation Maximization

The Expectation Maximization(EM) algorithm estimates the parameters of the multivariate probability density function in the form of a Gaussian mixture distribution with a specified number of mixtures.

Consider the set of the N feature vectors { x_1, x_2,...,x_{N} } from a d-dimensional Euclidean space drawn from a Gaussian mixture:

p(x;a_k,S_k, \pi _k) =  \sum _{k=1}^{m} \pi _kp_k(x),  \quad \pi _k  \geq 0,  \quad \sum _{k=1}^{m} \pi _k=1,

p_k(x)= \varphi (x;a_k,S_k)= \frac{1}{(2\pi)^{d/2}\mid{S_k}\mid^{1/2}} exp \left \{ - \frac{1}{2} (x-a_k)^TS_k^{-1}(x-a_k) \right \} ,

where m is the number of mixtures, p_k is the normal distribution density with the mean a_k and covariance matrix S_k, \pi_k is the weight of the k-th mixture. Given the number of mixtures M and the samples x_i, i=1..N the algorithm finds the maximum-likelihood estimates (MLE) of all the mixture parameters, that is, a_k, S_k and \pi_k :

L(x, \theta )=logp(x, \theta )= \sum _{i=1}^{N}log \left ( \sum _{k=1}^{m} \pi _kp_k(x) \right ) \to \max _{ \theta \in \Theta },

\Theta = \left \{ (a_k,S_k, \pi _k): a_k  \in \mathbbm{R} ^d,S_k=S_k^T>0,S_k  \in \mathbbm{R} ^{d  \times d}, \pi _k \geq 0, \sum _{k=1}^{m} \pi _k=1 \right \} .

The EM algorithm is an iterative procedure. Each iteration includes two steps. At the first step (Expectation step or E-step), you find a probability p_{i,k} (denoted \alpha_{i,k} in the formula below) of sample i to belong to mixture k using the currently available mixture parameter estimates:

\alpha _{ki} =  \frac{\pi_k\varphi(x;a_k,S_k)}{\sum\limits_{j=1}^{m}\pi_j\varphi(x;a_j,S_j)} .

At the second step (Maximization step or M-step), the mixture parameter estimates are refined using the computed probabilities:

\pi _k= \frac{1}{N} \sum _{i=1}^{N} \alpha _{ki},  \quad a_k= \frac{\sum\limits_{i=1}^{N}\alpha_{ki}x_i}{\sum\limits_{i=1}^{N}\alpha_{ki}} ,  \quad S_k= \frac{\sum\limits_{i=1}^{N}\alpha_{ki}(x_i-a_k)(x_i-a_k)^T}{\sum\limits_{i=1}^{N}\alpha_{ki}}

Alternatively, the algorithm may start with the M-step when the initial values for p_{i,k} can be provided. Another alternative when p_{i,k} are unknown is to use a simpler clustering algorithm to pre-cluster the input samples and thus obtain initial p_{i,k} . Often (including macnine learning) the kmeans() algorithm is used for that purpose.

One of the main problems of the EM algorithm is a large number of parameters to estimate. The majority of the parameters reside in covariance matrices, which are d \times d elements each where d is the feature space dimensionality. However, in many practical problems, the covariance matrices are close to diagonal or even to \mu_k*I , where I is an identity matrix and \mu_k is a mixture-dependent “scale” parameter. So, a robust computation scheme could start with harder constraints on the covariance matrices and then use the estimated parameters as an input for a less constrained optimization problem (often a diagonal covariance matrix is already a good enough approximation).


  • Bilmes98 J. A. Bilmes. A Gentle Tutorial of the EM Algorithm and its Application to Parameter Estimation for Gaussian Mixture and Hidden Markov Models. Technical Report TR-97-021, International Computer Science Institute and Computer Science Division, University of California at Berkeley, April 1998.


class CvEMParams

Parameters of the EM algorithm. All parameters are public. You can initialize them by a constructor and then override some of them directly if you want.


The constructors

C++: CvEMParams::CvEMParams()
C++: CvEMParams::CvEMParams(int nclusters, int cov_mat_type=CvEM::COV_MAT_DIAGONAL, int start_step=CvEM::START_AUTO_STEP, CvTermCriteria term_crit=cvTermCriteria(CV_TERMCRIT_ITER+CV_TERMCRIT_EPS, 100, FLT_EPSILON), const CvMat* probs=0, const CvMat* weights=0, const CvMat* means=0, const CvMat** covs=0 )
  • nclusters – The number of mixture components in the gaussian mixture model. Some of EM implementation could determine the optimal number of mixtures within a specified value range, but that is not the case in ML yet.
  • cov_mat_type

    Constraint on covariance matrices which defines type of matrices. Possible values are:

    • CvEM::COV_MAT_SPHERICAL A scaled identity matrix \mu_k * I. There is the only parameter \mu_k to be estimated for earch matrix. The option may be used in special cases, when the constraint is relevant, or as a first step in the optimization (for example in case when the data is preprocessed with PCA). The results of such preliminary estimation may be passed again to the optimization procedure, this time with cov_mat_type=CvEM::COV_MAT_DIAGONAL.
    • CvEM::COV_MAT_DIAGONAL A diagonal matrix with positive diagonal elements. The number of free parameters is d for each matrix. This is most commonly used option yielding good estimation results.
    • CvEM::COV_MAT_GENERIC A symmetric positively defined matrix. The number of free parameters in each matrix is about d^2/2. It is not recommended to use this option, unless there is pretty accurate initial estimation of the parameters and/or a huge number of training samples.
  • start_step

    The start step of the EM algorithm:

    • CvEM::START_E_STEP Start with Expectation step. You need to provide means a_k of mixture components to use this option. Optionally you can pass weights \pi_k and covariance matrices S_k of mixture components.
    • CvEM::START_M_STEP Start with Maximization step. You need to provide initial probabilites p_{i,k} to use this option.
    • CvEM::START_AUTO_STEP Start with Expectation step. You need not provide any parameters because they will be estimated by the k-means algorithm.
  • term_crit – The termination criteria of the EM algorithm. The EM algorithm can be terminated by the number of iterations term_crit.max_iter (number of M-steps) or when relative change of likelihood logarithm is less than term_crit.epsilon.
  • probs – Initial probabilities p_{i,k} of sample i to belong to mixture component k. It is a floating-point matrix of nsamples \times nclusters size. It is used and must be not NULL only when start_step=CvEM::START_M_STEP.
  • weights – Initial weights \pi_k of mixture components. It is a floating-point vector with nclusters elements. It is used (if not NULL) only when start_step=CvEM::START_E_STEP.
  • means – Initial means a_k of mixture components. It is a floating-point matrix of nclusters \times dims size. It is used used and must be not NULL only when start_step=CvEM::START_E_STEP.
  • covs – Initial covariance matrices S_k of mixture components. Each of covariance matrices is a valid square floating-point matrix of dims \times dims size. It is used (if not NULL) only when start_step=CvEM::START_E_STEP.

The default constructor represents a rough rule-of-the-thumb:

CvEMParams() : nclusters(10), cov_mat_type(1/*CvEM::COV_MAT_DIAGONAL*/),
    start_step(0/*CvEM::START_AUTO_STEP*/), probs(0), weights(0), means(0), covs(0)
    term_crit=cvTermCriteria( CV_TERMCRIT_ITER+CV_TERMCRIT_EPS, 100, FLT_EPSILON );

With another contstructor it is possible to override a variety of parameters from a single number of mixtures (the only essential problem-dependent parameter) to initial values for the mixture parameters.


class CvEM

The class implements the EM algorithm as described in the beginning of this section.


Estimates the Gaussian mixture parameters from a sample set.

C++: void CvEM::train(const Mat& samples, const Mat& sample_idx=Mat(), CvEMParams params=CvEMParams(), Mat* labels=0 )
C++: bool CvEM::train(const CvMat* samples, const CvMat* sampleIdx=0, CvEMParams params=CvEMParams(), CvMat* labels=0 )
Python: cv2.EM.train(samples[, sampleIdx[, params]]) → retval, labels
  • samples – Samples from which the Gaussian mixture model will be estimated.
  • sample_idx – Mask of samples to use. All samples are used by default.
  • params – Parameters of the EM algorithm.
  • labels – The optional output “class label” for each sample: \texttt{labels}_i=\texttt{arg max}_k(p_{i,k}), i=1..N (indices of the most probable mixture component for each sample).

Unlike many of the ML models, EM is an unsupervised learning algorithm and it does not take responses (class labels or function values) as input. Instead, it computes the Maximum Likelihood Estimate of the Gaussian mixture parameters from an input sample set, stores all the parameters inside the structure: p_{i,k} in probs, a_k in means , S_k in covs[k], \pi_k in weights , and optionally computes the output “class label” for each sample: \texttt{labels}_i=\texttt{arg max}_k(p_{i,k}), i=1..N (indices of the most probable mixture component for each sample).

The trained model can be used further for prediction, just like any other classifier. The trained model is similar to the CvNormalBayesClassifier.

For an example of clustering random samples of the multi-Gaussian distribution using EM, see em.cpp sample in the OpenCV distribution.


Returns a mixture component index of a sample.

C++: float CvEM::predict(const Mat& sample, Mat* probs=0 ) const
C++: float CvEM::predict(const CvMat* sample, CvMat* probs) const
Python: cv2.EM.predict(sample) → retval, probs
  • sample – A sample for classification.
  • probs – If it is not null then the method will write posterior probabilities of each component given the sample data to this parameter.


Returns the number of mixture components M in the gaussian mixture model.

C++: int CvEM::getNClusters() const
C++: int CvEM::get_nclusters() const
Python: cv2.EM.getNClusters() → retval


Returns mixture means a_k.

C++: Mat CvEM::getMeans() const
C++: const CvMat* CvEM::get_means() const
Python: cv2.EM.getMeans() → means


Returns mixture covariance matrices S_k.

C++: void CvEM::getCovs(std::vector<cv::Mat>& covs) const
C++: const CvMat** CvEM::get_covs() const
Python: cv2.EM.getCovs([covs]) → covs


Returns mixture weights \pi_k.

C++: Mat CvEM::getWeights() const
C++: const CvMat* CvEM::get_weights() const
Python: cv2.EM.getWeights() → weights


Returns vectors of probabilities for each training sample.

C++: Mat CvEM::getProbs() const
C++: const CvMat* CvEM::get_probs() const
Python: cv2.EM.getProbs() → probs

For each training sample i (that have been passed to the constructor or to CvEM::train()) returns probabilites p_{i,k} to belong to a mixture component k.


Returns logarithm of likelihood.

C++: double CvEM::getLikelihood() const
C++: double CvEM::get_log_likelihood() const
Python: cv2.EM.getLikelihood() → likelihood


Returns difference between logarithm of likelihood on the last iteration and logarithm of likelihood on the previous iteration.

C++: double CvEM::getLikelihoodDelta() const
C++: double CvEM::get_log_likelihood_delta() const
Python: cv2.EM.getLikelihoodDelta() → likelihood delta


Writes used parameters of the EM algorithm to a file storage.

C++: void CvEM::write_params(CvFileStorage* fs) const
  • fs – A file storage where parameters will be written.


Reads parameters of the EM algorithm.

C++: void CvEM::read_params(CvFileStorage* fs, CvFileNode* node)
  • fs – A file storage with parameters of the EM algorithm.
  • node – The parent map. If it is NULL, the function searches a node with parameters in all the top-level nodes (streams), starting with the first one.

The function reads EM parameters from the specified file storage node. For example of clustering random samples of multi-Gaussian distribution using EM see em.cpp sample in OpenCV distribution.