# relnCollFilt.py - Latent Feature-based Collaborative Filtering # Python 3 code. Full documentation at http://artint.info/code/python/code.pdf # Artificial Intelligence: Foundations of Computational Agents # http://artint.info # Copyright David L Poole and Alan K Mackworth 2016. # This work is licensed under a Creative Commons # Attribution-NonCommercial-ShareAlike 4.0 International License. # See: http://creativecommons.org/licenses/by-nc-sa/4.0/deed.en_US. import random import matplotlib.pyplot as plt import urllib.request class CF_learner(object): def __init__(self, ratings, # list of (u,i,r,t) ratings step_size = 0.01, # gradient descent step size regularization = 0.1, # the weight for the regulization terms num_features = 10, # number of hidden features feature_range = 0.2 # feaures are initialized to be between # -feature_range and feature_range ): self.ratings = ratings self.step_size = step_size self.regularization = regularization self.num_features = num_features self.num_ratings = len(ratings) self.ave_rating = (sum([rating for (user,item,rating,timestamp) in ratings]) /self.num_ratings) self.set_users_that_rated = set([user for (user,item,rating,timestamp) in ratings]) self.set_items_rated = set([item for (user,item,rating,timestamp) in ratings]) self.max_item_index = max(self.set_items_rated) self.max_user_index = max(self.set_users_that_rated) self.user_offset = [0]*(self.max_user_index+1) self.item_offset = [0]*(self.max_item_index+1) self.self_offset = 0 self.user_feature = [[random.uniform(-feature_range,feature_range) for f in range(num_features)] for i in range(self.max_user_index+1)] self.item_feature = [[random.uniform(-feature_range,feature_range) for f in range(num_features)] for i in range(self.max_item_index+1)] self.iter=0 def learn(self, num_iter = 50, # number of iterations of gradient descent trace = 1): """ do num_it iterations of gradient descent.""" for i in range(num_iter): self.iter += 1 total_error=0 sumsq_error=0 for (user,item,rating,timestamp) in self.ratings: prediction = (self.ave_rating + self.user_offset[user] + self.item_offset[item] + (self.self_offset if timestamp else 0) + sum([self.user_feature[user][f]*self.item_feature[item][f] for f in range(self.num_features)])) error = prediction - rating total_error += abs(error) sumsq_error += error * error if timestamp: self.self_offset += -self.step_size*error - self.step_size*self.regularization* self.self_offset self.user_offset[user] += ( -self.step_size*error -self.step_size*self.regularization* self.user_offset[user]) self.item_offset[item] += ( -self.step_size*error -self.step_size*self.regularization*self.item_offset[item]) for f in range(self.num_features): user_f = self.user_feature[user][f] item_f = self.item_feature[item][f] self.user_feature[user][f] += ( -self.step_size*error*item_f -self.step_size*self.regularization * user_f) self.item_feature[item][f] += ( -self.step_size*error*user_f -self.step_size*self.regularization*item_f) if trace>0: print("Iteration",i, "Ave Abs Error =",total_error/self.num_ratings, "Ave sq error =",sumsq_error/self.num_ratings) def plot_feature(self, f, # feature plot_all=False, # true if all points should be plotted num_points=200 # number of random points plotted if not all ): """plot some of the user-movie ratings, if plot_all is true num_points is the number of points selected at random plotted. the plot has the users on the x-axis sorted by their value on feature f and with the items on the y-axis sorted by their value on feature f and the ratings plotted at the corresponding x-y position. """ plt.ion() plt.xlabel("users") plt.ylabel("items") user_vals = [self.user_feature[user][f] for user in range(self.max_user_index+1) if user in self.set_users_that_rated] item_vals = [self.item_feature[item][f] for item in range(self.max_item_index+1) if item in self.set_items_rated] plt.axis([min(user_vals)-0.02, max(user_vals)+0.05, min(item_vals)-0.02, max(item_vals)+0.05]) if plot_all: for (user,item,rating,timestamp) in self.ratings: plt.text(self.user_feature[user][f], self.item_feature[item][f], str(rating),color='red' if timestamp else 'black') else: for i in range(num_points): (user,item,rating,timestamp) = random.choice(self.ratings) plt.text(self.user_feature[user][f], self.item_feature[item][f], str(rating), color='red' if timestamp else 'black') plt.show() class Rating_set(object): def __init__(self, local_file=False, url="http://files.grouplens.org/datasets/movielens/ml-100k/u.data", file_name="u.data", trace=1): if trace>0: print("reading...") if local_file: lines = open(file_name,'r') self.all_ratings = [tuple(int(e) for e in line.strip().split('\t')) for line in lines] else: lines = urllib.request.urlopen(url) self.all_ratings = [tuple(int(e) for e in line.decode('utf-8').strip().split('\t')) for line in lines] if trace>0: print("...read") def create_top_subset(self, num_items = 30, num_users = 30): """Returns a subset of the ratings by picking the most rated items, and then the users that have most ratings on these, and then all of the ratings that involve these users and items. """ max_item_index = max(item for (user,item,rating,timestamp) in self.all_ratings) item_counts = [0]*(max_item_index+1) for (user,item,rating,timestamp) in self.all_ratings: item_counts[item] += 1 items_sorted = sorted((item_counts[item],item) for item in range(max_item_index)) top_items = items_sorted[-num_items:] set_top_items = set(item for (count, item) in top_items) max_user_index = max(user for (user,item,rating,timestamp) in self.all_ratings) user_counts = [0]*(max_user_index+1) for (user,item,rating,timestamp) in self.all_ratings: if item in set_top_items: user_counts[user] += 1 users_sorted = sorted((user_counts[user],user) for user in range(max_user_index)) top_users = users_sorted[-num_users:] set_top_users = set(user for (count, user) in top_users) used_ratings = [ (user,item,rating,timestamp) for (user,item,rating,timestamp) in self.all_ratings if user in set_top_users and item in set_top_items] return used_ratings #movielens = Rating_set(local_file=True) #learner0 = CF_learner(movielens.all_ratings) #learner0.learn(20) # full dataset of 100,000 ratings #learner0.plot_feature(0) # try other features too #movielens_subset = movielens.create_top_subset(num_items = 20, num_users = 20) #learner1 = CF_learner(movielens_subset, num_features=1) #learner1.learn(100) #smaller dataset of around 400 ratings #learner1.plot_feature(0,plot_all=True) #This is the analysis of Assignmet 1 grades. #Tuples are (student, page, mark, own_page) #from cs522 import cs522as1 #learner522 = CF_learner(cs522as1, regularization = 0, num_features=1) # learner522.learn(100) # try this a few times # learner522.plot_feature(0,plot_all=True) # sorted(learner522.item_offset) # sorted(learner522.user_offset) # learner522.self_offset