def mds(matrix,n,outfile):
'''Code for multi-dimensional scaling
Take the eigendecomp of the gram matrix.
Take the p largest eigenvectors where
rank(G)< p,
? p should be bounded by the number of attributes?
? so p < 3?
'''
print "--- Starting Multi-Dimensional Scaling ---\n"
gram = gram_matrix(matrix,n)
print "--- Starting Eigen Decomp --- \n"
matrix_w = eigen_decomp(gram,dim= n)
return matrix_w
def lap(k,data,style_data,dim,outfile):
''' 1. Construct your k nearest neighbors adjacency matrix
2. Use Heat Equation exp -{||x_i-x_j||^2/4t}
'''
adjacency = knn(data,data.shape[1])
heat = 1.0
print "--- Seeting Weights on Adjacency Matrix ---\n"
heat_matrix = set_weights(adjacency,k,heat)
print "--- Creating Laplacian Matrix ---\n"
laplacian, weight = get_laplacian(heat_matrix)
print "--- Eigendecomp -> Points ---\n"
points = eigen_decomp( laplacian,weight,dim)
print "--- Plotting Points ---\n"
plot_points(points,style_data, outfile)