Help protect the Great Barrier Reef with TensorFlow on Kaggle Join Challenge


View source on GitHub

Dirichlet-Multinomial compound distribution.

Inherits From: Distribution

The Dirichlet-Multinomial distribution is parameterized by a (batch of) length-K concentration vectors (K > 1) and a total_count number of trials, i.e., the number of trials per draw from the DirichletMultinomial. It is defined over a (batch of) length-K vector counts such that tf.reduce_sum(counts, -1) = total_count. The Dirichlet-Multinomial is identically the Beta-Binomial distribution when K = 2.

Mathematical Details

The Dirichlet-Multinomial is a distribution over K-class counts, i.e., a length-K vector of non-negative integer counts = n = [n_0, ..., n_{K-1}].

The probability mass function (pmf) is,

pmf(n; alpha, N) = Beta(alpha + n) / (prod_j n_j!) / Z
Z = Beta(alpha) / N!


  • concentration = alpha = [alpha_0, ..., alpha_{K-1}], alpha_j > 0,
  • total_count = N, N a positive integer,
  • N! is N factorial, and,
  • Beta(x) = prod_j Gamma(x_j) / Gamma(sum_j x_j) is the multivariate beta function, and,
  • Gamma is the gamma function.

Dirichlet-Multinomial is a compound distribution, i.e., its samples are generated as follows.

  1. Choose class probabilities: probs = [p_0,...,p_{K-1}] ~ Dir(concentration)
  2. Draw integers: counts = [n_0,...,n_{K-1}] ~ Multinomial(total_count, probs)

The last concentration dimension parametrizes a single Dirichlet-Multinomial distribution. When calling distribution functions (e.g., dist.prob(counts)), concentration, total_count and counts are broadcast to the same shape. The last dimension of counts corresponds single Dirichlet-Multinomial distributions.

Distribution parameters are automatically broadcast in all functions; see examples for details.


The number of classes, K, must not exceed:

  • the largest integer representable by self.dtype, i.e., 2**(mantissa_bits+1) (IEE754),
  • the maximum Tensor index, i.e., 2**31-1.

In other words,

K <= min(2**31-1, {
  tf.float16: 2**11,
  tf.float32: 2**24,
  tf.float64: 2**53 }[param.dtype])


alpha = [1., 2., 3.]
n = 2.
dist = DirichletMultinomial(n, alpha)

Creates a 3-class distribution, with the 3rd class is most likely to be drawn. The distribution functions can be evaluated on counts.

# counts same shape as alpha.
counts = [0., 0., 2.]
dist.prob(counts)  # Shape []

# alpha will be broadcast to [[1., 2., 3.], [1., 2., 3.]] to match counts.
counts = [[1., 1., 0.], [1., 0., 1.]]
dist.prob(counts)  # Shape [2]

# alpha will be broadcast to shape [5, 7, 3] to match counts.
counts = [[...]]  # Shape [5, 7, 3]
dist.prob(counts)  # Shape [5, 7]

Creates a 2-batch of 3-class distributions.

alpha = [[1., 2., 3.], [4., 5., 6.]]  # Shape [2, 3]
n = [3., 3.]
dist = DirichletMultinomial(n, alpha)

# counts will be broadcast to [[2., 1., 0.], [2., 1., 0.]] to match alpha.
counts = [2., 1., 0.]
dist.prob(counts)  # Shape [2]

total_count Non-negative floating point tensor, whose dtype is the same as concentration. The shape is broadcastable to [N1,..., Nm] with m >= 0. Defines this as a batch of N1 x ... x Nm different Dirichlet multinomial distributions. Its components should be equal to integer values.
concentration Positive floating point tensor, whose dtype is the same as n with shape broadcastable to [N1,..., Nm, K] m >= 0. Defines this as a batch of N1 x ... x Nm different K class Dirichlet multinomial distributions.
validate_args Python bool, default False. When True distribution parameters are checked for validity despite possibly degrading runtime performance. When False invalid inputs may silently render incorrect outputs.
allow_nan_stats Python bool, default True. When True, statistics (e.g., mean, mode, variance) use the value "NaN" to indicate the result is undefined. When False, an exception is raised if one or more of the statistic's batch members are undefined.
name Python str name prefixed to Ops created by this class.

allow_nan_stats Python bool describing behavior when a stat is undefined.

Stats return +/- infinity when it makes sense. E.g., the variance of a Cauchy distribution is infinity. However, sometimes the statistic is undefined, e.g., if a distribution's pdf does not achieve a maximum within the support of the distribution, the mode is undefined. If the mean is undefined, then by definition the variance is undefined. E.g. the mean for Student's T for df = 1 is undefined (no clear way to say it is either + or - infinity), so the variance = E[(X - mean)**2] is also undefined.

batch_shape Shape of a single sample from a single event index as a TensorShape.

May be partially defined or unknown.

The batch dimensions are indexes into independent, non-identical parameterizations of this distribution.

concentration Concentration parameter; expected prior counts for that coordinate.
dtype The DType of Tensors handled by this Distribution.
event_shape Shape of a single sample from a single batch as a TensorShape.

May be partially defined or unknown.

name Name prepended to all ops created by this Distribution.
parameters Dictionary of parameters used to instantiate this Distribution.
reparameterization_type Describes how samples from the distribution are reparameterized.

Currently this is one of the static instances distributions.FULLY_REPARAMETERIZED or distributions.NOT_REPARAMETERIZED.

total_concentration Sum of last dim of concentration parameter.
total_count Number of trials used to construct a sample.
validate_args Python bool indicating possibly expensive checks are enabled.



View source

Shape of a single sample from a single event index as a 1-D Tensor.

The batch dimensions are indexes into independent, non-identical parameterizations of this distribution.

name name to give to the op

batch_shape Tensor.


View source

Cumulative distribution function.

Given random variable X, the cumulative distribution function cdf is:

cdf(x) := P[X <= x]

value float or double Tensor.
name Python str prepended to names of ops created by this function.

cdf a Tensor of shape sample_shape(x) + self.batch_shape with values of type self.dtype.


View source

Creates a deep copy of the distribution.

**override_parameters_kwargs String/value dictionary of initialization arguments to override with new values.

distribution A new instance of type(self) initialized from the union of self.parameters and override_parameters_kwargs, i.e., dict(self.parameters, **override_parameters_kwargs).


View source


Covariance is (possibly) defined only for non-scalar-event distributions.

For example, for a length-k, vector-valued distribution, it is calculated as,

Cov[i, j] = Covariance(X_i, X_j) = E[(X_i - E[X_i]) (X_j - E[X_j])]

where Cov is a (batch of) k x k matrix, 0 <= (i, j) < k, and E denotes expectation.

Alternatively, for non-vector, multivariate distributions (e.g., matrix-valued, Wishart), Covariance shall return a (batch of) matrices under some vectorization of the events, i.e.,

Cov[i, j] = Covariance(Vec(X)_i, Vec(X)_j) = [as above]

where Cov is a (batch of) k' x k' matrices, 0 <= (i, j) < k' = reduce_prod(event_shape), and Vec is some function mapping indices of this distribution's event dimensions to indices of a length-k' vector.

Additional documentation from DirichletMultinomial:

The covariance for each batch member is defined as the following:

Var(X_j) = n * alpha_j / alpha_0 * (1 - alpha_j / alpha_0) *
(n + alpha_0) / (1 + alpha_0)

where concentration = alpha and total_concentration = alpha_0 = sum_j alpha_j.

The covariance between elements in a batch is defined as:

Cov(X_i, X_j) = -n * alpha_i * alpha_j / alpha_0 ** 2 *
(n + alpha_0) / (1 + alpha_0)

name Python str prepended to names of ops created by this function.

covariance Floating-point Tensor with shape [B1, ..., Bn, k', k'] where the first n dimensions are batch coordinates and k' = reduce_prod(self.event_shape).


View source

Computes the (Shannon) cross entropy.

Denote this distribution (self) by P and the other distribution by Q. Assuming P, Q are absolutely continuous with respect to one another and permit densities p(x) dr(x) and q(x) dr(x), (Shanon) cross entropy is defined as:

H[P, Q] = E_p[-log q(X)] = -int_F p(x) log q(x) dr(x)

where F denotes the support of the random variable X ~ P.

other tfp.distributions.Distribution instance.
name Python str prepended to names of ops created by this function.