Rasoul Karimi

Towards Optimal Active Learning for Matrix Factorization in Recommender Systems

Recommender systems help web users to address information overload. However their performance depends on the number of provided ratings by users. This problem is amplified for a new user because he/she has not provided any rating. To address this problem, active learning methods have been proposed to acquire those ratings from users, that will help most in determining their interests. The optimal active learning selects a query that directly optimizes the expected error for the test data. This approach is applicable for prediction models in which this question can be answered in closed-form given the distribution of test data is known. Unfortunately, there are many tasks and models for which the optimal selection cannot efficiently be found in closed-form. Therefore, most of the active learning methods optimize different, non-optimal criteria, such as uncertainty. Nevertheless, in this paper we exploit the characteristics of matrix factorization, which leads to a closed-form solution and by being inspired from existing optimal active learning for the regression task, develop a method that approximates the optimal solution for recommender systems. Our results demonstrate that the proposed method improves the prediction accuracy of MF.

Non-myopic active learning for recommender systems based on Matrix Factorization

Recommender systems help Web users to address information overload. However, their performance depends on the number of provided ratings by users. This problem is amplified for a new user because he/she has not provided any ratings. In this paper, we consider the new user problem as an optimization problem and propose a non-myopic active learning method to select items to be queried from the new user. The proposed method is based on Matrix Factorization (MF) which is a strong prediction model for recommender systems. First, the proposed method explores the latent space to get closer to the optimal new user parameters. Then, it exploits the learned parameters and slightly adjusts them. The results show that beside improving the accuracy of recommendation, MF approach also results in drastically reduced user waiting times, i.e., the time that the users wait before being asked a new query. Therefore, it is an ideal choice for using active learning in real-world applications of recommender systems.

Active learning for aspect model in recommender systems

Recommender systems help Web users to address information overload. Their performance, however, depends on the amount of information that users provide about their preferences. Users are not willing to provide information for a large amount of items, thus the quality of recommendations is affected specially for new users. Active learning has been proposed in the past, to acquire preference information from users. Based on an underlying prediction model, these approaches determine the most informative item for querying the new user to provide a rating. In this paper, we propose a new active learning method which is developed specially based on aspect model features. There is a difference between classic active learning and active learning for recommender system. In the recommender system context, each item has already been rated by training users while in classic active learning there is not training user. We take into account this difference and develop a new method which competes with a complicated bayesian approach in accuracy while results in drastically reduced (one order of magnitude) user waiting times, i.e., the time that the users wait before being asked a new query.

A supervised active learning framework for recommender systems based on decision trees

A key challenge in recommender systems is how to profile new users. A well-known solution for this problem is to ask new users to rate a few items to reveal their preferences and to use active learning to find optimally informative items. Compared to the application of active learning in classification (regression), active learning in recommender systems presents several differences: although there are no ratings for new users, there is an abundance of available ratings—collectively—from past (existing) users. In this paper, we propose an innovative approach for active learning in recommender systems, which aims at taking advantage of this additional information. The main idea is to consider existing users as (hypothetical) new users and solve an active learning problem for each of them. In the end, we aggregate all solved problems in order to learn how to solve the active learning problem for a real new user. As the ratings of existing users (i.e., labels) are known and are used for active learning purposes, the proposed framework is in fact a supervised active learning framework. Based on this framework, we investigate two different types of models: the first model is based on information about average item ratings and the second on matrix factorization. We present experimental results on the Netflix dataset, which show that the proposed approach significantly outperforms state-of-the-art baselines.

Factorized Decision Trees for Active Learning in Recommender Systems

A key challenge in recommender systems is how to profile new users. A well-known solution for this problem is to use active learning techniques and ask the new user to rate a few items to reveal her preferences. The sequence of queries should not be static, i.e in each step the best query depends on the responses of the new user to the previous queries. Decision trees have been proposed to capture the dynamic aspect of this process. In this paper we improve decision trees in two ways. First, we propose the Most Popular Sampling (MPS) method to increase the speed of the tree construction. In each node, instead of checking all candidate items, only those which are popular among users associated with the node are examined. Second, we develop a new algorithm to build decision trees. It is called Factorized Decision Trees (FDT) and exploits matrix factorization to predict the ratings at nodes of the tree. The experimental results on the Netflix dataset show that both contributions are successful. The MPS increases the speed of the tree construction without harming the accuracy. And FDT improves the accuracy of rating predictions especially in the last queries.

RFID-Enhanced Museum for Interactive Experience

Visitors to physical museums are often overwhelmed by the vast amount of information available in the space they are exploring, making it difficult to select personally interesting content. Personalization solutions can provide the required user-centered interactivity between the visitors and the museum websites or museum guide systems. Recommender systems are among the most successful personalization technologies, as they have already been incorporated to solve similar problems in e-commerce, where users have a lot of choices to select a product. However, developing recommender system for museums is more challenging, because in contrast to e-commerce, museums and their exhibits exist in a physical world. Therefore, we need a hardware technology to provide us the required infrastructure to observe and model the environment and user activities. Radio-frequency identification (RFID) technology is among the best solutions for this issue, because it is cheap, fast, robust, and available everywhere. In this paper, we describe the vision of our project called RFID-Enhanced Museum for Interactive Experience (REMIX), which aims to developing a personalization platform for museums based on RFID technology and advanced recommender-systems algorithms.

Active Learning for Recommender Systems

Recommender systems learn user preferences and provide them personalized recommendations. Evidently, the performance of recommender systems depends on the amount of information that users provide regarding items, most often in the form of ratings. This problem is amplified for new users because they have not provided any rating, which impacts negatively on the quality of generated recommendations. This problem is called new-user problem. A simple and effective way to overcome this problem is posing queries to new users so that they express their preferences about selected items, e.g., by rating them. Nevertheless, the selection of items must take into consideration that users are not willing to answer a lot of such queries. To address this problem, active learning methods have been proposed to acquire the most informative ratings, i.e., ratings from users that will help most in determining their interests. Active learning is a learning algorithm that is able to interactively query the Oracle to obtain labels for data instances. The Oracle is a user or teacher who knows the labels. The aim of this dissertation [8] is to take inspiration from the literature of active learning for classification (regression) problems and develop new methods for the new-user problem in recommender systems. In the recommender system context, new users play the role of the Oracle and provide ratings (labels) to items (data instances). Specifically, the following questions are addressed in this dissertation: (1) which recommendation model is suitable for active-learning purposes? (Sect. 2) (2) how can active learning criteria be adapted and customized for the new-user problem and which one is the best? (Sect. 3) (3) what are the specific requirements and properties of the new-user problem that do not exist in active learning and how can new active learning methods be developed based on these properties? (Sects. 4, 5).