From beca8098ad92f3b1e1a84b62fb75629685318c6f Mon Sep 17 00:00:00 2001
From: hellboy <j.moudrik@gmail.com>
Date: Sat, 27 Feb 2010 17:06:59 +0100
Subject: [PATCH] knn.py ADDED: k-NearestNeighbour algorithm for vector
 classification.

---
 knn.py | 116 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 116 insertions(+)
 create mode 100755 knn.py

diff --git a/knn.py b/knn.py
new file mode 100755
index 0000000..8c2e923
--- /dev/null
+++ b/knn.py
@@ -0,0 +1,116 @@
+#!/usr/bin/python
+import sys
+from gostyle import *
+from math import sqrt
+
+from data_about_players import Data
+
+class KNNOutputVectorGenerator(VectorGenerator):
+	""" k-NearestNeighbour output vector generator."""
+	def __init__(self, ref_dict, k=2):
+		"""
+			ref_dict is a dictionary of refence input/output vectors.
+			e.g. ref_dict= { (1.0,2.0):(9.0,16.0,21.0)
+		"""
+		self.ref_dict = ref_dict
+		self.k = k
+	def __call__(self, player_vector):
+		distance=[]
+		for ref_vec in ref_dict.keys():
+			distance.append((self.distance(ref_vec, player_vector), ref_vec))
+		distance.sort()	
+		
+		#print "DBG :"
+		#print distance
+
+		ref_output_vecs = [ self.ref_dict[b] for a,b in distance[:self.k] ]
+		coefs = [ self.weight_fc(a) for a,b in distance[:self.k] ]
+
+		return linear_combination(ref_output_vecs, coefs)
+	def weight_fc(self, distance):
+		return 0.2 ** distance
+	def distance(self, vec1, vec2):
+		if len(vec1) != len(vec2):
+			raise RuntimeError("Dimensions of vectors mismatch.")
+		return sqrt(sum([ (float(a) - float(b))**2 for a,b in zip(vec1,vec2)]))
+		
+
+if __name__ == '__main__':
+	main_pat_filename = Data.main_pat_filename
+	filename_play_other = 'knn_other.data'
+	filename_play_ref = 'knn_ref.data'
+	filename_play_ref_orig = 'knn_ref_orig.data'
+	num_features = 300
+	players_all = Data.players_all
+	players_ref = Data.player_vector.keys()
+	players_other =  [ x for x in players_all if x not in players_ref ]
+
+	### Object creating input vector when called
+	print >>sys.stderr, "Creating input vector generator from main pat file:", main_pat_filename
+	i = InputVectorGenerator(main_pat_filename, num_features)
+
+	# Create list of input vectors
+	input_vectors_ref = []
+	for name in players_ref: 
+		input_vectors_ref += [i(Data.pat_files_folder + name)]
+	input_vectors_other = []
+	for name in players_other: 
+		input_vectors_other += [i(Data.pat_files_folder + name)]
+
+	if len(input_vectors_ref) == 0:
+		print >>sys.stderr, "No reference vectors."
+		sys.exit()
+	if len(input_vectors_other) == 0:
+		print >>sys.stderr, "No vectors to process."
+		sys.exit()
+
+	### PCA example usage
+	# Change this to False, if you do not want to use PCA
+	use_pca = False
+	if use_pca:
+		# Create PCA object, trained on input_vectors
+		print >>sys.stderr, "Running PCA."
+		pca = PCA(input_vectors_ref + input_vectors_other, reduce=True)
+		# Perform a PCA on input vectors
+		input_vectors_ref = pca.process_list_of_vectors(input_vectors_ref)
+		input_vectors_other = pca.process_list_of_vectors(input_vectors_other)
+		# Creates a Composed object that first generates an input vector
+		# and then performs a PCA analysis on it.
+		i = Compose(i, pca)
+
+	### Object creating output vector when called;
+	ref_dict = {}
+	for name, input_vector in zip(players_ref, input_vectors_ref):
+		ref_dict[tuple(input_vector)] = Data.player_vector[name]
+	
+	oknn = KNNOutputVectorGenerator(ref_dict, k=5)
+
+	
+	# Create list of output vectors using weighted kNN algorithm approximating output_vector
+	output_vectors_other = [ oknn(input_vector) for input_vector in input_vectors_other ]
+	output_vectors_ref = [ oknn(input_vector) for input_vector in input_vectors_ref ]
+
+	def print_me( names, vecs, where):
+		if len(names) != len(vecs):
+			raise RuntimeError("Dimensions of vectors mismatch.")
+			
+		f = open(where, 'w')
+		print >>sys.stderr, "Saving output_vectors to file:", where
+
+		for i in xrange(len(names)):
+			name_to_print = '_'.join(names[i].split())
+			print_vector([name_to_print] + list(vecs[i]), f)
+
+		f.close()
+	
+	print_me(players_ref, output_vectors_ref, filename_play_ref)
+	print_me(players_other, output_vectors_other, filename_play_other)
+
+	f = open(filename_play_ref_orig, 'w')
+	print >>sys.stderr, "Saving output_vectors to file:", filename_play_ref_orig
+	for name, vec in Data.player_vector.items():
+		name_to_print = '_'.join(name.split())
+		print_vector([name_to_print]+vec, f)
+		
+	f.close()
+
-- 
2.11.4.GIT