reshaping data properly mesh grid

Hi all,

so I have the follwoing problem. I have data stored in a (2000,4) mattrix where the last colum represents the predictions I drew from a logistic regression. Now, I managed to put in a messhgrid data on columns 2 and 3 (first term is the constant) as height and weeightt as follows:

h = min(X(:,1)):.01:max(X(:,1)); %(1,2069)
w = min(X(:,2)):1:max(X(:,2)); %(1,24)
[hx, wx] = meshgrid(h,w); %(24,2069)

but did not manage to put in the same form the predictions since I computed in a later step. Predictions are a (2000,1) vector.

The aim basically is to make this:

pred = reshape(pred, size(hx));

work without the error:

Error using reshape
Number of elements must not change. Use [] as one of the size inputs to automatically calculate the appropriate size
for that dimension.

which, if I have undestood the issue is solved if I manage to put pred in a vector of size "size(hx, 1) * size(hx, 2)".

Can you pleasse help m out on this?

Answer 1

[KSSV]

Let X be your data matrix of size 2000*4.

h = X(:,2) ;
w = X(:,3) ;
nh = length(unique(h)) ; 
nw = length(unique(w)) ; 
h = reshape(h,nh,nw) ; 
w = reshape(w,nh,nw) ; 
p = reshape(X(:,4),nh,nw) ; 

If the above works, it is good. If not have a look on griddata function. If not attach your data.

[federico nutarelli]

@KSSV thank you for your reply. AActually h and w are not the econd and third columns of X but are derived from X as follows (dimensions are commented):

h = min(X(:,1)):.01:max(X(:,1)); %(1,2069)
w = min(X(:,2)):1:max(X(:,2)); %(1,24)
[hx, wx] = meshgrid(h,w); %(24,2069)

I am actually trying o reproduce Lab:SVM of this website. In particular I have done the logistic function and obtained the prediction called "pred", but cannot reshape them properly

[KSSV]

h = min(X(:,1)):.01:max(X(:,1)); %(1,2069)
w = min(X(:,2)):1:max(X(:,2)); %(1,24)
[hx, wx] = meshgrid(h,w); %(24,2069)
pred = griddata(X(:,1),X(:,2),X(:,4),h,w) ; 

[federico nutarelli]

It raiss the error:

Error using griddata (line 106)
XI and YI must be the same size or vectors of different orientations.

Actually: X(:,1),X(:,2),X(:,4) have all the same dimension (2000,1). But h and w have different dimensions (as above)

[federico nutarelli]

@KSSV I attached the zip of th entir foldr. I am using main.m. From line 44 to line 73 I am applying logistic regression. he problem arises at lines 108-109, where I should reshape data to construct a countourf()

[KSSV]

You need to use:

pred = griddata(X(:,1),X(:,2),pred,hx,wx) ;

[federico nutarelli]

It works!