Error using randsample (line 89) W must have length 3.

function [fuzzy_rules, pheromone] = aco_fuzzy_rules(X, y, num_rules, num_ants, max_iter, alpha, beta, rho, Q)

num_features = size(X, 2);

num_membership_functions = 3; % Assuming 3 membership functions per feature

% Convert table to numeric array

X = table2array(X);

% Initialize pheromone matrix

pheromone = ones(num_features, num_membership_functions * num_rules);

% Initialize best solution

best_solution = [];

best_fitness = -inf;

for iter = 1:max_iter

% Construct solutions (fuzzy rules) using ants

solutions = cell(num_ants, 1);

for k = 1:num_ants

solution = [];

for j = 1:num_features

p = pheromone(j, :) .^ alpha .* (1 ./ (1 + abs(X(:, j)))) .^ beta;

p = p / sum(p);

if length(p) < num_membership_functions

p = [p, zeros(1, num_membership_functions - length(p))];

end

for r = 1:num_rules

solution = [solution, randsample(1:num_membership_functions, 1, true, p')];

end

end

solutions{k} = solution;

end

% Evaluate fitness of solutions

fitnesses = cellfun(@(s) fuzzy_fitness(s, X, y), solutions);

% Update best solution

[best_fitness, best_idx] = max([best_fitness, fitnesses]);

best_solution = solutions{best_idx};

% Update pheromone matrix

pheromone = (1 - rho) * pheromone;

for k = 1:num_ants

solution = solutions{k};

fitness = fitnesses(k);

pheromone = pheromone + Q * fitness * (solution == reshape(repmat(1:num_membership_functions, num_rules, 1), 1, []));

end

end

% Convert best solution to fuzzy rules

fuzzy_rules = reshape(best_solution, num_features, num_rules);

end

The above code is for ACO to be used to learn new rules

can someone help how to fix this error - Error using randsample (line 89)

W must have length 3.

Error in AcoGafuzzyNew>aco_fuzzy_rules (line 103)

solution = [solution, randsample(1:num_membership_functions, 1,

true, p')];

Error in AcoGafuzzyNew (line 16)

[fuzzy_rules, pheromone] = aco_fuzzy_rules(X, y, num_rules, num_ants, max_iter, alpha, beta, rho, Q);

Answer 1

[Balavignesh]

Hi Michael,

It looks like you are encountering an error with 'randsample' MATLAB function. The error suggests that the probability weights vector 'p' you are passing doesn't have the same length as 'num_membership_functions'. This inconsistency in length likely arises during the calculation of 'p', especially if the 'pheromone' matrix dimensions or its manipulation does not align perfectly with the expected sizes.

for k = 1:num_ants
    solution = [];
    for j = 1:num_features
        p = pheromone(j, :) .^ alpha .* (1 ./ (1 + abs(X(:, j)))) .^ beta;
        p = p / sum(p);
        
        % Ensure p has the correct length by resizing it if necessary
        expected_length = num_membership_functions * num_rules;
        if length(p) < expected_length
            p = [p, zeros(1, expected_length - length(p))];
        elseif length(p) > expected_length
            p = p(1:expected_length);
        end
        
        % Now generate rules for each feature across all rules
        for r = 1:num_rules
            idx_start = (r-1)*num_membership_functions + 1;
            idx_end = r*num_membership_functions;
            rule_p = p(idx_start:idx_end); % Extract the probabilities for the current rule
            solution = [solution, randsample(1:num_membership_functions, 1, true, rule_p)];
        end
    end
    solutions{k} = solution;
end

In this adjusted code snippet, I've added a step to ensure that 'p' is resized to match the expected length 'expected_length'. This length is calculated based on the number of membership functions and the number of rules, ensuring that p aligns with the structure you're working with. When selecting a membership function for each rule, it's important to correctly partition the probability vector of the correct length of 'num_membership_functions', which should resolve the error you encountered in my opinion.

Kindly have a look at the following documentation links to have more information on:

• 'randsample' function: https://www.mathworks.com/help/stats/randsample.html

Hope that helps!

Balavignesh