I have 100 matrices in which each row corresponds to an individual and column refers to sites. I want to sort the row by a measure of similarity such that the most similar individuals are next to each other in a matrix. I used k-nearest neighbours to sort the matrices by rows and I give these sorted matrices to a convolutional neural network. I want to know if there are other measures by which I can achieve the task in hand. The code I use for k-nearest neighbour is:
def sort_min_diff(amat):
mb = NearestNeighbors(len(amat), metric='manhattan').fit(amat)
v = mb.kneighbors(amat)
smallest = np.argmin(v[0].sum(axis=1))
return amat[v[1][smallest]]
X_snp = np.array(snp_matrix)
q = []
for i in range(len(X_snp)):
q.append((sort_min_diff(X_snp[i])))
q = np.array(q)
My X_snp matrix is of shape (100,60,4500) that is I have 100 such matrices. Also, my matrices are filled with 0 and 1. Suggestions would be appreciated.
