computeCompleteDataLogLikelihood

[ghost]

Can anyone explain the following function ? What is the function based on?

float computeCompleteDataLogLikelihood(const ResidualIterator& first_residual, const ResidualIterator& last_residual, const WeightIterator& first_weight, const Eigen::Vector2f& mean, const Eigen::Matrix2f& precision)
{
size_t n = (last_residual - first_residual);
size_t c = 1;
double error_sum = 0.0;
double error_acc = 1.0;

for(ResidualIterator err_it = first_residual; err_it != last_residual; ++err_it, ++c)
{
error_acc *= (1.0 + 0.2 * ((*err_it).transpose() * precision * (*err_it))(0,0));

if((c % 50) == 0)
{
  error_sum += std::log(error_acc);
  error_acc = 1.0;
}

}

return 0.5 * n * std::log(precision.determinant()) - 0.5 * (5.0 + 2.0) * error_sum;
}

[MarcoLomb]

Hi,
did you find an answer by yourself? I have the same doubt and I can't understand why we have to refresh the error_acc every 50 iterations