blob d53009599ae27c6f533c20ce82b90b6659ce3d68
1 #!/usr/bin/python
2 import sys
3 from gostyle import *
4 from math import sqrt
8 class KNNOutputVectorGenerator(VectorGenerator):
9 """ k-NearestNeighbour output vector generator."""
10 def __init__(self, ref_dict, k=2):
11 """
12 ref_dict is a dictionary of refence input/output vectors.
13 e.g. ref_dict= { (1.0,2.0):(9.0,16.0,21.0)
14 """
15 self.ref_dict = ref_dict
16 self.k = k
17 def __call__(self, player_vector):
18 distance=[]
19 for ref_vec in ref_dict.keys():
20 distance.append((self.distance(ref_vec, player_vector), ref_vec))
21 distance.sort()
23 ref_output_vecs = [ self.ref_dict[b] for a,b in distance[:self.k] ]
24 coefs = [ self.weight_fc(a) for a,b in distance[:self.k] ]
26 return linear_combination(ref_output_vecs, coefs)
27 def weight_fc(self, distance):
28 return 0.9 ** distance
29 def distance(self, vec1, vec2):
30 if len(vec1) != len(vec2):
31 raise RuntimeError("Dimensions of vectors mismatch.")
32 return sqrt(sum([ (float(a) - float(b))**2 for a,b in zip(vec1,vec2)]))
35 if __name__ == '__main__':
36 main_pat_filename = Data.main_pat_filename
37 filename_play_other = 'knn_other.data'
38 filename_play_ref = 'knn_ref.data'
39 filename_play_ref_orig = 'knn_ref_orig.data'
40 num_features = 300
41 k = 4
42 players_all = Data.players_all
43 players_ref = Data.player_vector.keys()
44 players_other = [ x for x in players_all if x not in players_ref ]
46 ### Object creating input vector when called
47 print >>sys.stderr, "Creating input vector generator from main pat file:", main_pat_filename
48 i = InputVectorGenerator(main_pat_filename, num_features)
50 # Create list of input vectors
51 input_vectors_ref = []
52 for name in players_ref:
53 input_vectors_ref += [i(Data.pat_files_folder + name)]
54 input_vectors_other = []
55 for name in players_other:
56 input_vectors_other += [i(Data.pat_files_folder + name)]
58 if len(input_vectors_ref) == 0:
59 print >>sys.stderr, "No reference vectors."
60 sys.exit()
61 if len(input_vectors_other) == 0:
62 print >>sys.stderr, "No vectors to process."
63 sys.exit()
65 ### PCA example usage
66 # Change this to False, if you do not want to use PCA
67 use_pca = False
68 if use_pca:
69 # Create PCA object, trained on input_vectors
70 print >>sys.stderr, "Running PCA."
71 pca = PCA(input_vectors_ref + input_vectors_other, reduce=True)
72 # Perform a PCA on input vectors
73 input_vectors_ref = pca.process_list_of_vectors(input_vectors_ref)
74 input_vectors_other = pca.process_list_of_vectors(input_vectors_other)
75 # Creates a Composed object that first generates an input vector
76 # and then performs a PCA analysis on it.
77 i = Compose(i, pca)
78 #print input_vectors_other[0]
79 ### Object creating output vector when called;
80 ref_dict = {}
81 for name, input_vector in zip(players_ref, input_vectors_ref):
82 ref_dict[tuple(input_vector)] = Data.player_vector[name]
84 oknn = KNNOutputVectorGenerator(ref_dict, k=k)
86 # Create list of output vectors using weighted kNN algorithm approximating output_vector
87 output_vectors_other = [ oknn(input_vector) for input_vector in input_vectors_other ]
88 output_vectors_ref = [ oknn(input_vector) for input_vector in input_vectors_ref ]
90 def print_me( names, vecs, where):
91 if len(names) != len(vecs):
92 raise RuntimeError("Dimensions of vectors mismatch.")
94 f = open(where, 'w')
95 print >>sys.stderr, "Saving output_vectors to file:", where
97 for i in xrange(len(names)):
98 name_to_print = '_'.join(names[i].split())
99 print_vector([name_to_print] + list(vecs[i]), f)
101 f.close()
103 print_me(players_ref, [Data.player_vector[name] for name in players_ref], filename_play_ref_orig)
104 print_me(players_ref, output_vectors_ref, str(k)+filename_play_ref)
105 print_me(players_other, output_vectors_other, str(k)+filename_play_other)