Codice: Seleziona tutto
sudo apt-get install octavePoiché octave 4.0.3 del repo di Debian Stretch restituisce un grafico correttamente 'ingrigliato', gradirei sapere:
Codice: Seleziona tutto
1.. come fare a sradicare da Ubuntu, senza lasciar tracce residue, l'attuale v. 4.4.2
2.. come configurare e installare correttamnte la v. 4.0.3 di octave, reperibile
da qui: https://ftp.gnu.org/gnu/octave/ [old versions]
Codice: Seleziona tutto
% ------- file bfit.m ----------
% Avviare con: octave bfit.m
% ------------------------------
%% Bayesian Curve Fitting
% Author: Karl Ezra Pilario, 31 Aug 2018
% Reference: Bishop, Pattern Recognition and Machine Learning, 2006
%% Generate training data and set parameters
clear;
N = 15; % Number of data points
x = linspace(0,1,N); % Training data, x
t = sin(2*pi*x) + 0.2*randn(size(x)); % Training data, t
X = linspace(-0.5,1.5,100); % Define finely separated x values
T = sin(2*pi*X); % Compute a smooth curve for X
% Set the constant values. Note: precision = 1/variance
a = 5e-3; % alpha, precision of the weights, w
b = 11.1; % beta, precision of the inputs, x
D = 8; % degree of polynomial to be fitted
M = D+1; % no. of parameters in the model
W = 0.5*[1 -1]; % no. of std devs for predictive distribution
% Randomize incoming training data points
ind = randperm(N);
t = t(ind); x = x(ind);
clc;
for j = 1:N
%% Perform sequential curve fitting
t2 = t(1:j); x2 = x(1:j); % Collect points 1 to j
phi = bsxfun(@power,x2',0:D)'; % Design matrix, Eq. (3.16)
Sinv = a*eye(M) + b*(phi*phi'); % Cov(weights), Eq. (1.72)
PHI = bsxfun(@power,X',0:D)'; % Finely separated x
s = zeros(size(X)); m = s;
for k = 1:length(X)
m(k) = b*PHI(:,k)'*(Sinv\(phi*t2')); % Mean, Eq. (1.70)
s(k) = 1/b + PHI(:,k)'*(Sinv\PHI(:,k)); % Var, Eq. (1.71)
end
%% Plot the results
f = figure(1);
clf; scatter(x2,t2,'b','filled'); % Plot data points 1 to j
hold on; grid on; box on;
axis([-0.1 1.1 -1.5 1.5]); % Fix the axes
plot(X,T,'g',X,m,'r-','LineWidth',2); % Plot m(x)
p1 = W(1)*sqrt(s)+m; % Solve for +W*s(x)
p2 = W(2)*sqrt(s)+m; % Solve for -W*s(x)
p = [X fliplr(X); p1 fliplr(p2)]; % Boundary pts for fill
h = fill(p(1,:),p(2,:),'m'); % Plot +/-W*s(x)
set(h,'facealpha',0.2); % Make fill translucent
title(sprintf('%d out of %d data points',j,N));
legend({'Observed Data','Ground Truth',...
'Mean Estimate','Distribution Estimate'},...
'Location','southoutside');
print(f, '-dpng', '-r180', num2str(j))
% disp('Press any key to continue...'); pause; clc;
end
Salve, facciamo un po' d'ordine.
Ciao