Revision 332b50d164b504cdb8f766c171cf147458a0b074 authored by Vincent ADAM on 18 February 2019, 16:24:13 UTC, committed by Vincent ADAM on 18 February 2019, 16:24:13 UTC
1 parent b0fb55d
Raw File
kullback_leiblers.py
# Copyright 2016 James Hensman, alexggmatthews
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

# -*- coding: utf-8 -*-


import tensorflow as tf

from . import settings
from .decors import name_scope


@name_scope()
def gauss_kl(q_mu, q_sqrt, K=None):
    """
    Compute the KL divergence KL[q || p] between

          q(x) = N(q_mu, q_sqrt^2)
    and
          p(x) = N(0, K)

    We assume N multiple independent distributions, given by the columns of
    q_mu and the last dimension of q_sqrt. Returns the sum of the divergences.

    q_mu is a matrix (M x L), each column contains a mean.

    q_sqrt can be a 3D tensor (L x M x M), each matrix within is a lower
        triangular square-root matrix of the covariance of q.
    q_sqrt can be a matrix (M x L), each column represents the diagonal of a
        square-root matrix of the covariance of q.

    K is the covariance of p.
    It is a positive definite matrix (M x M) or a tensor of stacked such matrices (L x M x M)
    If K is None, compute the KL divergence to p(x) = N(0, I) instead.
    """

    white = K is None
    diag = q_sqrt.get_shape().ndims == 2

    M, B = tf.shape(q_mu)[0], tf.shape(q_mu)[1]

    if white:
        alpha = q_mu  # M x B
    else:
        batch = K.get_shape().ndims == 3

        Lp = tf.cholesky(K)  # B x M x M or M x M
        q_mu = tf.transpose(q_mu)[:, :, None] if batch else q_mu  # B x M x 1 or M x B
        alpha = tf.matrix_triangular_solve(Lp, q_mu, lower=True)  # B x M x 1 or M x B

    if diag:
        Lq = Lq_diag = q_sqrt
        Lq_full = tf.matrix_diag(tf.transpose(q_sqrt))  # B x M x M
    else:
        Lq = Lq_full = tf.matrix_band_part(q_sqrt, -1, 0)  # force lower triangle # B x M x M
        Lq_diag = tf.matrix_diag_part(Lq)  # M x B

    # Mahalanobis term: μqᵀ Σp⁻¹ μq
    mahalanobis = tf.reduce_sum(tf.square(alpha))

    # Constant term: - B * M
    constant = - tf.cast(tf.size(q_mu, out_type=tf.int64), dtype=settings.float_type)

    # Log-determinant of the covariance of q(x):
    logdet_qcov = tf.reduce_sum(tf.log(tf.square(Lq_diag)))

    # Trace term: tr(Σp⁻¹ Σq)
    if white:
        trace = tf.reduce_sum(tf.square(Lq))
    else:
        if diag and not batch:
            # K is M x M and q_sqrt is M x B: fast specialisation
            LpT = tf.transpose(Lp)  # M x M
            Lp_inv = tf.matrix_triangular_solve(Lp, tf.eye(M, dtype=settings.float_type),lower=True)  # M x M
            K_inv = tf.matrix_diag_part(tf.matrix_triangular_solve(LpT, Lp_inv, lower=False))[:, None]  # M x M -> M x 1
            trace = tf.reduce_sum(K_inv * tf.square(q_sqrt))
        else:
            # TODO: broadcast instead of tile when tf allows (not implemented in tf <= 1.6.0)
            Lp_full = Lp if batch else tf.tile(tf.expand_dims(Lp, 0), [B, 1, 1])
            LpiLq = tf.matrix_triangular_solve(Lp_full, Lq_full, lower=True)
            trace = tf.reduce_sum(tf.square(LpiLq))

    twoKL = mahalanobis + constant - logdet_qcov + trace

    # Log-determinant of the covariance of p(x):
    if not white:
        log_sqdiag_Lp = tf.log(tf.square(tf.matrix_diag_part(Lp)))
        sum_log_sqdiag_Lp = tf.reduce_sum(log_sqdiag_Lp)
        # If K is B x M x M, num_latent is no longer implicit, no need to multiply the single kernel logdet
        scale = 1.0 if batch else tf.cast(B, settings.float_type)
        twoKL += scale * sum_log_sqdiag_Lp

    return 0.5 * twoKL
back to top