Compute the derivative of the negative log probability with respect to the parameters of one spot, given some samples of the other spots. More...

Inheritance diagram for SpotNegProbabilityDiffWithSampledBackground:

Public Member Functions
	SpotNegProbabilityDiffWithSampledBackground (const SampledBackgroundData &d)
Vector< 4 >	operator() (const Vector< 4 > &spot) const
Public Attributes
const vector< vector< vector < double > > > &	sample_intensities_without_spot
const vector< vector< double > > &	pixel_intensities
const vector< ImageRef >	pixels
double	mu_brightness
double	sigma_brightness
double	mu_blur
double	sigma_blur
const Matrix< 3 >	A
const Vector< 3 >	pi
double	variance
const vector< int >	O

Detailed Description

Compute the derivative of the negative log probability with respect to the parameters of one spot, given some samples of the other spots.

Definition at line 567 of file multispot5.cc.

Constructor & Destructor Documentation

SpotNegProbabilityDiffWithSampledBackground::SpotNegProbabilityDiffWithSampledBackground ( const SampledBackgroundData & d ) [inline]

Parameters:

d	Necessary data for construction

Definition at line 570 of file multispot5.cc.

    :SampledBackgroundData(d)
    {
    }

Member Function Documentation

Vector<4> SpotNegProbabilityDiffWithSampledBackground::operator() ( const Vector< 4 > & spot ) const [inline]

Compute the probability of spot.

Parameters:

spot	Spot position

Definition at line 577 of file multispot5.cc.

References SampledMultispot::compute_spot_intensity_derivatives(), diff_log_log_normal(), and forward_algorithm_deriv().

    {
        if(spot[0] <= 0 || spot[1] <= 0)
            return Ones * std::numeric_limits<double>::quiet_NaN();

        vector<pair<double, Vector<4> > > spot_intensities = compute_spot_intensity_derivatives(pixels, spot);

        Vector<4> sum_diff_log = Zeros;

        for(unsigned int s=0; s < sample_intensities_without_spot.size(); s++)
        {
            SpotWithBackground B(sample_intensities_without_spot[s], spot_intensities, pixel_intensities, variance);

            pair<double, Vector<4> > r = forward_algorithm_deriv(A, pi, B, O);
            
            sum_diff_log += r.second;
        }

        Vector<4> diff_log = sum_diff_log / sample_intensities_without_spot.size();

        //Compute the log probability of the prior
        Vector<4> logprior_deriv = makeVector(diff_log_log_normal(spot[0], mu_brightness, sigma_brightness),
                                              diff_log_log_normal(spot[1], mu_blur, sigma_blur), 0, 0);
        
        return -(diff_log + logprior_deriv);
    }