Problem with Question 8

[adrwend]

@fstella I have problems with accurately applying the bayesian decoding to the 2D simulation data. I have identified the problem: xy_sampling is 2D now (x + y dimension), it needs to be compared to a 2D xy_decoded. The problem is, I don't understand how to decode each dimension individually so that xy_decoded has different values in both dimensions. I suspect to use different dimensions of the 2D estimated firing rate two calculate probabilities for x and y dimenstion individually, but I am unsure on how to do that. This is how I currently do it (I was not able to attach MatLab files to this Issue) :

% Identify precision via bayesian decoding and calculating median error
% estimated_firing_rate2D has more dimenstions than spikes2D, hence for comparison they need to be aligned:

aligned_estim_firing_rate2D = reshape(estim_firing_rate2D, [], n_cells2D ); % I am not sure whether this reshaping is correct

xy_decoded=zeros(size(spikes2D,2),2); % This needs to be 2D, too

for t_bin = 1:size(spikes2D,2)
    
    if(sum(spikes2D(:,t_bin),1)>0) % Check if the time window contains spikes
    
    Post_p2D=NaN(size(aligned_estim_firing_rate2D,2),2); % Column dimension 2 for now
    
    for i=1:size(aligned_estim_firing_rate2D,1)
        
        % Note that we work with log so that we can sum probabilities instead of multiplying them
        % Here I need to somehow calculate probabilities for 2 dimensions individually:
        Post_p2D(i,1)=sum(log(poisspdf(spikes2D(:,t_bin),aligned_estim_firing_rate2D(:,i)/sampling_rate2D))); 
        Post_p2D(i,2)=
        
    end
          
    xy_decoded(t_bin,:) = find(Post_p2D==nanmax(Post_p2D),1,'first');

    else
        
    xy_decoded(t_bin,:) = NaN;   
    
    end
end

bayesian_mederr2D = nanmean(nanmedian(sqrt(((xy_sampling-xy_decoded*bin_size2D).^2))))
% The resulting median error coming from this code is 400+, probably because both columns of xy_decoded have identical values.

Any help would be appreciated.