Filters








1,874 Hits in 4.4 sec

Online collaborative filtering with nearly optimal dynamic regret

Baruch Awerbuch, Thomas P. Hayes
2007 Proceedings of the nineteenth annual ACM symposium on Parallel algorithms and architectures - SPAA '07  
As a metric of success for this problem, we consider dynamic regret, i.e., regret versus the off-line optimal sequence of decisions.  ...  We present an algorithm for this setting whose expected dynamic regret per honest player is optimal up to a multiplicative constant and an additive polylogarithmic term, assuming the number of options  ...  However, with multiple trustworthy agents, it may be possible in some scenarios to achieve low regret against dynamic optimal strategies. Our goal is to study regret against dynamic strategies.  ... 
doi:10.1145/1248377.1248431 dblp:conf/spaa/AwerbuchH07 fatcat:pytummk5yvbjlpq47e45jmznjy

BanditMF: Multi-Armed Bandit Based Matrix Factorization Recommender System [article]

Shenghao Xu
2021 arXiv   pre-print
For collaborative filtering, the classical method is training the model offline, then perform the online testing, but this approach can no longer handle the dynamic changes in user preferences which is  ...  of scarcity of valid information, (2) how to solve the sub-optimal problem of bandit algorithms in strong social relations domains caused by independently estimating unknown parameters associated with  ...  For collaborative filtering, the classical method is training the model offline, then perform the online testing, but this approach can no longer handle the dynamic changes in user preferences which is  ... 
arXiv:2106.10898v2 fatcat:upnm5pmva5agvenkcuolwv5muq

Context-Aware Online Learning for Course Recommendation of MOOC Big Data [article]

Yifan Hou, Pan Zhou, Ting Wang, Li Yu, Yuchong Hu, Dapeng Wu
2016 arXiv   pre-print
In this paper, we propose a big data-supported, context-aware online learning-based course recommender system that could handle the dynamic and infinitely massive datasets, which recommends courses by  ...  The Massive Open Online Course (MOOC) has expanded significantly in recent years.  ...  Apropos of filtering-based approaches, there are some branches such like collaborative filtering [10] [13], content-based filtering [14] and hybrid approaches [15] [16] .  ... 
arXiv:1610.03147v2 fatcat:bslk4r37pvcqbdkgebzi6qpzxi

Differentially Private Online Learning for Cloud-Based Video Recommendation with Multimedia Big Data in Social Networks [article]

Pan Zhou, Yingxue Zhou, Dapeng Wu, Hai Jin
2016 arXiv   pre-print
With the rapid growth in multimedia services and the enormous offers of video contents in online social networks, users have difficulty in obtaining their interests.  ...  To handle the problems, we propose a cloud-assisted differentially private video recommendation system based on distributed online learning.  ...  ACKNOWLEDGMENT This research is supported by National Science Foundation of China with Grant 61401169.  ... 
arXiv:1509.00181v7 fatcat:lye3lamanjehpiwffx4aenfyfq

Feedback Adaptive Learning for Medical and Educational Application Recommendation

Cem Tekin, Sepehr Elahi, Mihaela Van Der Schaar
2020 IEEE Transactions on Services Computing  
episodic versions of ✏n-greedy, Thompson sampling, and collaborative filtering methods.  ...  Then, we propose our online learning algorithm, named FeedBack Adaptive Learning (FeedBAL), and prove that its regret with respect to the benchmark increases logarithmically in expectation.  ...  ACKNOWLEDGMENTS The authors would like to thank Ken Cheung and Xinyu Hu of Columbia University for providing us with IntelliCare data that was used in the simulations.  ... 
doi:10.1109/tsc.2020.3037224 fatcat:53memzm4uzhfreidmyc3whw4am

Robot fast adaptation to changes in human engagement during simulated dynamic social interaction with active exploration in parameterized reinforcement learning

Mehdi Khamassi, George Velentzas, Theodore Tsitsimis, Costas Tzafestas
2018 IEEE Transactions on Cognitive and Developmental Systems  
be also enhanced with the use of sibling Kalman filters.  ...  The diagonal terms of the covariance matrix COV in the Kalman filter nearly monotonically decrease, resulting in a large variance σ t when action a 6 is executed until about timestep 600, and progressively  ... 
doi:10.1109/tcds.2018.2843122 fatcat:5c64vbuft5dklhotrt2kkg5w7q

An algorithm for online tensor prediction [article]

John Pothier, Josh Girson, Shuchin Aeron
2015 arXiv   pre-print
Then following a similar construction as in [3], we exploit this algorithm to propose an online algorithm for learning and prediction of tensors with provable regret guarantees.  ...  The result indicate superior performance compared to other (online) convex tensor completion methods.  ...  In particular, the shadow that propagates across the city can be seen as a "trend" emerging and fading in a time-dynamic collaborative filtering setting. C.  ... 
arXiv:1507.07974v1 fatcat:yegora5u5relfgyx7sqjq3z5da

Forecasting the nearly unforecastable: why aren't airline bookings adhering to the prediction algorithm?

Saravanan Thirumuruganathan, Soon-gyo Jung, Dianne Ramirez Robillos, Joni Salminen, Bernard J. Jansen
2021 Electronic Commerce Research  
We then compare the performance of the Next Likely Destination model in a real-life consumer study with 35,000 actual airline customers.  ...  Acknowledgements We thank the international airline company for its collaboration in this research.  ...  The overall predictive power was 23%, with collaborative filtering having a predictive power of 30% and 19% for the ensemble method.  ... 
doi:10.1007/s10660-021-09457-0 fatcat:5zin2rli5nfnnoetq4fwjg3klq

Competitive Collaborative Learning [chapter]

Baruch Awerbuch, Robert D. Kleinberg
2005 Lecture Notes in Computer Science  
This bound compares favorably with the naïve approach in which each user ignores feedback from peers and chooses resources using a multi-armed bandit algorithm; in this case the expected regret per user  ...  However, it is also clear that the presence of malicious, dishonest users in the community threatens the usefulness of such collaborative learning processes.  ...  Theoretical work on collaborative filtering has mostly dealt with centralized algorithms for such problems.  ... 
doi:10.1007/11503415_16 fatcat:jfqmob4cuzgepariqldlu2meuu

Competitive collaborative learning

Baruch Awerbuch, Robert Kleinberg
2008 Journal of computer and system sciences (Print)  
This bound compares favorably with the naïve approach in which each user ignores feedback from peers and chooses resources using a multi-armed bandit algorithm; in this case the expected regret per user  ...  However, it is also clear that the presence of malicious, dishonest users in the community threatens the usefulness of such collaborative learning processes.  ...  Theoretical work on collaborative filtering has mostly dealt with centralized algorithms for such problems.  ... 
doi:10.1016/j.jcss.2007.08.004 fatcat:eennsppqsjenzoyu75oul4mnwu

Preference Dynamics Under Personalized Recommendations

Sarah Dean, Jamie Morgenstern
2022 Proceedings of the 23rd ACM Conference on Economics and Computation  
Polarization in particular can occur even in ecosystems with "mass media, " where no personalization takes place, as recently explored in a natural model of preference dynamics by Hązła et al. [14] and  ...  and Google The design of content recommendation systems underpins many online platforms: social media feeds, online news aggregators, and audio/video hosting websites all choose how best to organize an  ...  ACKNOWLEDGMENTS This work was supported by an NSF CAREER award (ID 2045402), An NSF AI Center Award (The Institute for Foundations of Machine Learning), and the Simons Collaboration on the Theory of Algorithmic  ... 
doi:10.1145/3490486.3538346 fatcat:4ryddh35p5fs5jbda77n35e66y

Fighting Boredom in Recommender Systems with Linear Reinforcement Learning

Romain Warlop, Alessandro Lazaric, Jérémie Mary
2018 Neural Information Processing Systems  
We then introduce an extension of the UCRL algorithm (LINUCRL) to effectively balance exploration and exploitation in an unknown environment, and we derive a regret bound that is independent of the number  ...  Unlike in the general online learning scenario, in our setting the transition function f is known and thus the regret incurred from bad estimates of the dynamics is reduced to zero.  ...  In practice, one may prefer to use collaborative filtering algorithms (e.g., matrix factorisation) and apply our proposed algorithm on top of them to find the optimal cadence to maximize long term performances  ... 
dblp:conf/nips/WarlopLM18 fatcat:57q5c5qdarah7m5gz673oezv2m

Case-studies on exploiting explicit customer requirements in recommender systems

Markus Zanker, Markus Jessenitschnig
2008 User modeling and user-adapted interaction  
Its contribution lies in comparing different techniques such as knowledge-and utility-based methods, collaborative filtering, association rule mining as well as hybrid variants when user models consist  ...  Recommender Systems (RS) suggest useful and interesting items to users in order to increase user satisfaction and online conversion rates.  ...  In nearly all collaborative recommender systems similarity between users is based on ratings.  ... 
doi:10.1007/s11257-008-9048-y fatcat:52ppiafdorcnthyq5sbg3yvlzm

Toward Optimal Adversarial Policies in the Multiplicative Learning System with a Malicious Expert [article]

S. Rasoul Etesami, Negar Kiyavash, Vincent Leon, H. Vincent Poor
2020 arXiv   pre-print
For the online setting where the malicious expert can adaptively make its decisions, we show that the optimal online policy can be efficiently computed by solving a dynamic program in O(N^3).  ...  with an approximation ratio of 1+O(√(ln N/N)), where N is the total number of prediction stages.  ...  Moreover, almost all cases of collaborative filtering or distributed decision making are vulnerable to such internal threats.  ... 
arXiv:2001.00543v2 fatcat:asbk4ly4mncslkha4n6fxuvm4m

Certified Defenses for Data Poisoning Attacks [article]

Jacob Steinhardt, Pang Wei Koh, Percy Liang
2017 arXiv   pre-print
Our bound comes paired with a candidate attack that often nearly matches the upper bound, giving us a powerful tool for quickly assessing defenses on a given dataset.  ...  Machine learning systems trained on user-provided data are susceptible to data poisoning attacks, whereby malicious users inject false training data with the aim of corrupting the learned model.  ...  average regret Regret( n) n is small will have a nearly optimal candidate attack D p .  ... 
arXiv:1706.03691v2 fatcat:je7sdqmgcrcpjmimp2vco7rmgi
« Previous Showing results 1 — 15 out of 1,874 results