sequential_importance_particle_filter.m 5.17 KB
Newer Older
1
function [LIK,lik] = sequential_importance_particle_filter(ReducedForm,Y,start,ParticleOptions,ThreadsOptions)
Stéphane Adjemian's avatar
Stéphane Adjemian committed
2

3
4
% Evaluates the likelihood of a nonlinear model with a particle filter (optionally with resampling).

Stéphane Adjemian's avatar
Stéphane Adjemian committed
5
% Copyright (C) 2011-2014 Dynare Team
6
%
Stéphane Adjemian's avatar
Stéphane Adjemian committed
7
% This file is part of Dynare (particles module).
8
9
10
11
12
13
%
% Dynare is free software: you can redistribute it and/or modify
% it under the terms of the GNU General Public License as published by
% the Free Software Foundation, either version 3 of the License, or
% (at your option) any later version.
%
Stéphane Adjemian's avatar
Stéphane Adjemian committed
14
% Dynare particles module is distributed in the hope that it will be useful,
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
% but WITHOUT ANY WARRANTY; without even the implied warranty of
% MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
% GNU General Public License for more details.
%
% You should have received a copy of the GNU General Public License
% along with Dynare.  If not, see <http://www.gnu.org/licenses/>.

persistent init_flag
persistent mf0 mf1
persistent number_of_particles number_of_state_variables
persistent sample_size number_of_observed_variables number_of_structural_innovations

% Set default value for start
if isempty(start)
    start = 1;
end

% Set flag for prunning
33
pruning = ParticleOptions.pruning;
34
35
36
37
38
39
40
41
42
43
44
45
46
47

% Get steady state and mean.
steadystate = ReducedForm.steadystate;
constant = ReducedForm.constant;
state_variables_steady_state = ReducedForm.state_variables_steady_state;

% Set persistent variables (if needed).
if isempty(init_flag)
    mf0 = ReducedForm.mf0;
    mf1 = ReducedForm.mf1;
    sample_size = size(Y,2);
    number_of_state_variables = length(mf0);
    number_of_observed_variables = length(mf1);
    number_of_structural_innovations = length(ReducedForm.Q);
48
    number_of_particles = ParticleOptions.number_of_particles;
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
    init_flag = 1;
end

% Set local state space model (first order approximation).
ghx  = ReducedForm.ghx;
ghu  = ReducedForm.ghu;

% Set local state space model (second order approximation).
ghxx = ReducedForm.ghxx;
ghuu = ReducedForm.ghuu;
ghxu = ReducedForm.ghxu;

% Get covariance matrices.
Q = ReducedForm.Q; % Covariance matrix of the structural innovations.
H = ReducedForm.H; % Covariance matrix of the measurement errors.
if isempty(H)
    H = 0;
end

% Initialization of the likelihood.
const_lik = log(2*pi)*number_of_observed_variables;
lik  = NaN(sample_size,1);

% Get initial condition for the state vector.
StateVectorMean = ReducedForm.StateVectorMean;
StateVectorVarianceSquareRoot = reduced_rank_cholesky(ReducedForm.StateVectorVariance)';
if pruning
    StateVectorMean_ = StateVectorMean;
    StateVectorVarianceSquareRoot_ = StateVectorVarianceSquareRoot;
end

% Get the rank of StateVectorVarianceSquareRoot
state_variance_rank = size(StateVectorVarianceSquareRoot,2);

% Factorize the covariance matrix of the structural innovations
Q_lower_triangular_cholesky = chol(Q)';

% Set seed for randn().
set_dynare_seed('default');

% Initialization of the weights across particles.
weights = ones(1,number_of_particles)/number_of_particles ;
StateVectors = bsxfun(@plus,StateVectorVarianceSquareRoot*randn(state_variance_rank,number_of_particles),StateVectorMean);
if pruning
    StateVectors_ = StateVectors;
end

% Loop over observations
for t=1:sample_size
    yhat = bsxfun(@minus,StateVectors,state_variables_steady_state);
    epsilon = Q_lower_triangular_cholesky*randn(number_of_structural_innovations,number_of_particles);
    if pruning
        yhat_ = bsxfun(@minus,StateVectors_,state_variables_steady_state);
102
        [tmp, tmp_] = local_state_space_iteration_2(yhat,epsilon,ghx,ghu,constant,ghxx,ghuu,ghxu,yhat_,steadystate,ThreadsOptions.local_state_space_iteration_2);
103
    else
104
        tmp = local_state_space_iteration_2(yhat,epsilon,ghx,ghu,constant,ghxx,ghuu,ghxu,ThreadsOptions.local_state_space_iteration_2);
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
    end
    PredictedObservedMean = tmp(mf1,:)*transpose(weights);
    PredictionError = bsxfun(@minus,Y(:,t),tmp(mf1,:));
    dPredictedObservedMean = bsxfun(@minus,tmp(mf1,:),PredictedObservedMean);
    PredictedObservedVariance = bsxfun(@times,dPredictedObservedMean,weights)*dPredictedObservedMean' + H;
    if rcond(PredictedObservedVariance) > 1e-16
        lnw = -.5*(const_lik+log(det(PredictedObservedVariance))+sum(PredictionError.*(PredictedObservedVariance\PredictionError),1));
    else
        LIK = NaN;
        return
    end
    dfac = max(lnw);
    wtilde = weights.*exp(lnw-dfac);
    lik(t) = log(sum(wtilde))+dfac;
    weights = wtilde/sum(wtilde);
120
    if (ParticleOptions.resampling.status.generic && neff(weights)<ParticleOptions.resampling.threshold*sample_size) || ParticleOptions.resampling.status.systematic
121
        if pruning
122
            temp = resample([tmp(mf0,:)' tmp_(mf0,:)'],weights',ParticleOptions);
123
124
125
            StateVectors = temp(:,1:number_of_state_variables)';
            StateVectors_ = temp(:,number_of_state_variables+1:2*number_of_state_variables)';
        else
126
            StateVectors = resample(tmp(mf0,:)',weights',ParticleOptions)';
127
128
        end
        weights = ones(1,number_of_particles)/number_of_particles;
129
    elseif ParticleOptions.resampling.status.none
130
131
132
133
134
135
136
137
        StateVectors = tmp(mf0,:);
        if pruning
            StateVectors_ = tmp_(mf0,:);
        end
    end
end

LIK = -sum(lik(start:end));