Misspecified Gaussian Process Bandits
Dec 6, 2021
Speakers
Ilija Bogunovic
Speaker · 0 followers
Andreas Krause
Speaker · 6 followers
About
We consider the problem of optimizing a black-box function based on noisy bandit feedback. Kernelized bandit algorithms have shown strong empirical and theoretical performance for this problem. They heavily rely on the assumption that the model is well-specified, however, and can fail without it. Instead, we introduce and address a misspecified kernelized bandit setting where the unknown function can be ϵ–uniformly approximated by a function with a bounded norm in some Reproducing Kernel Hilbert Space (RKHS). We design efficient and practical algorithms whose performance degrades minimally in the presence of model misspecification. Specifically, we present two algorithms based on Gaussian process (GP) methods: an optimistic EC-GP-UCB algorithm that requires knowing the misspecification error, and Phased GP Uncertainty Sampling, an elimination-type algorithm that can adapt to unknown model misspecification. We provide upper bounds on their cumulative regret in terms of ϵ, the time horizon, and the underlying kernel, and we show that our algorithm achieves optimal dependence on ϵ with no prior knowledge of misspecification. In addition, in a stochastic contextual setting, we show that EC-GP-UCB can be effectively combined with the regret bound balancing strategy and attain similar regret bounds despite not knowing ϵ.We consider the problem of optimizing a black-box function based on noisy bandit feedback. Kernelized bandit algorithms have shown strong empirical and theoretical performance for this problem. They heavily rely on the assumption that the model is well-specified, however, and can fail without it. Instead, we introduce and address a misspecified kernelized bandit setting where the unknown function can be ϵ–uniformly approximated by a function with a bounded norm in some Reproducing Kernel Hilbert…
Organizer
NeurIPS 2021
Account · 1.9k followers
About NeurIPS 2021
Neural Information Processing Systems (NeurIPS) is a multi-track machine learning and computational neuroscience conference that includes invited talks, demonstrations, symposia and oral and poster presentations of refereed papers. Following the conference, there are workshops which provide a less formal setting.
Like the format? Trust SlidesLive to capture your next event!
Professional recording and live streaming, delivered globally.
Sharing
Recommended Videos
Presentations on similar topic, category or speaker
Neural optimal feedback control with local learning rules
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Which priors matter? Benchmarking models for learning latent dynamics
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
BNS: Building Network Structures Dynamically for Continual Learning
Watch later
Qi Qin, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Meaning in Context: Pragmatic Communication in Humans and Machines
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Orthogonal Quantile Regression
Watch later
Shai Feldman, …
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%
Permutation Compressors for Provably Faster Distributed Nonconvex Optimization
Watch later
Total of 0 viewers voted for saving the presentation to eternal vault which is 0.0%