Cautious R Regret M Minimization: Online O Optimization w with L Long-Term B Budg udget Co Constraints Nikolaos Liakopoulos 1,2,3 , Apostolos Destounis 1 , Georgios S. Paschos 1 , Thrasyvoulos Spyropoulos 2 , Panayotis Mertikopoulos 4 1 France Research Center, Huawei Technologies Co. Ltd. , 2 Eurecom, Mobile Communications Department 3 UPMC, Sorbonne Universities, 4 University of Grenoble Alpes, CNRS, Inria, LIG ICML 2019, Long Beach, 13 June 2019
Motivati ting E Example: le: On Online Ad e Adver erti tisem emen ent Cautious Regret Minimization: Online Optimization with Long-Term Budget Constraints, ICML 2019, Long Beach N. Liakopoulos, A. Destounis, G. Paschos, T. Spyropoulos, P. Mertikopoulos
Motivati ting E Example: le: On Online Ad e Adver erti tisem emen ent Select over d websites to place bids for advertisements , over slots up to a budget of • You select your bid and then pay the a priori unknown • Depends on: competitor fraud, website popularity change, customer behavior • Long term budget constraint • Return profit is revealed after the decision • Cautious Regret Minimization: Online Optimization with Long-Term Budget Constraints, ICML 2019, Long Beach N. Liakopoulos, A. Destounis, G. Paschos, T. Spyropoulos, P. Mertikopoulos
Motivati ting E Example: le: On Online Ad e Adver erti tisem emen ent Select over d websites to place bids for advertisements , over slots up to a budget of • You select your bid and then pay the a priori unknown • Depends on: competitor fraud, website popularity change, customer behavior • Long term budget constraint • Return profit is revealed after the decision • Target Online Optimization Cautious Regret Minimization: Online Optimization with Long-Term Budget Constraints, ICML 2019, Long Beach N. Liakopoulos, A. Destounis, G. Paschos, T. Spyropoulos, P. Mertikopoulos
Prior or W Work Cautious Regret Minimization: Online Optimization with Long-Term Budget Constraints, ICML 2019, Long Beach N. Liakopoulos, A. Destounis, G. Paschos, T. Spyropoulos, P. Mertikopoulos
Prior or W Work Online Convex Optimization with Long Term Adversarial Constraints Cautious Regret Minimization: Online Optimization with Long-Term Budget Constraints, ICML 2019, Long Beach N. Liakopoulos, A. Destounis, G. Paschos, T. Spyropoulos, P. Mertikopoulos
Prior or W Work Online Convex Optimization with Long Term Adversarial Constraints “No Regret” against the weakest benchmark Impossibility against the strongest benchmark Cautious Regret Minimization: Online Optimization with Long-Term Budget Constraints, ICML 2019, Long Beach N. Liakopoulos, A. Destounis, G. Paschos, T. Spyropoulos, P. Mertikopoulos
Cautious O Online Lagrang angian an Des esce cent Cautious Regret Minimization: Online Optimization with Long-Term Budget Constraints, ICML 2019, Long Beach N. Liakopoulos, A. Destounis, G. Paschos, T. Spyropoulos, P. Mertikopoulos
Cautious O Online Lagrang angian an Des esce cent K-Benchmark Cautious Regret Minimization: Online Optimization with Long-Term Budget Constraints, ICML 2019, Long Beach N. Liakopoulos, A. Destounis, G. Paschos, T. Spyropoulos, P. Mertikopoulos
Cautious O Online Lagrang angian an Des esce cent K-Benchmark COLD Cautious Regret Minimization: Online Optimization with Long-Term Budget Constraints, ICML 2019, Long Beach N. Liakopoulos, A. Destounis, G. Paschos, T. Spyropoulos, P. Mertikopoulos
Cautious O Online Lagrang angian an Des esce cent K-Benchmark Achievable Bounds [Yuan et al., Sun et al.] COLD Cautious Regret Minimization: Online Optimization with Long-Term Budget Constraints, ICML 2019, Long Beach N. Liakopoulos, A. Destounis, G. Paschos, T. Spyropoulos, P. Mertikopoulos
Impact o t of Gu Guarantee Cautious Regret Minimization: Online Optimization with Long-Term Budget Constraints, ICML 2019, Long Beach N. Liakopoulos, A. Destounis, G. Paschos, T. Spyropoulos, P. Mertikopoulos
Thanks! See you at our poster! Cautious Regret Minimization: Online Optimization with Long-Term Budget Constraints, ICML 2019, Long Beach N. Liakopoulos, A. Destounis, G. Paschos, T. Spyropoulos, P. Mertikopoulos
Recommend
More recommend