Prediction, learning, and games

By Nicolò Cesa-Bianchi, Gábor Lugosi. 2006

4,451 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Economics and Econometrics, Game & Decision Theory

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Economics and Econometrics

In Game & Decision Theory

Citations in the corpus

On the Complexity of Bandit and Derivative-Free Stochastic Convex Optimization (2012, background): “L G ] 2 9 A pr 2 01 powerful models for sequential decision making under uncertainty Cesa-Bianchi and Lugosi (2006); Bubeck and Cesa-Bianchi (2012).”
Many Paths to Equilibrium: GANs Do Not Need to Decrease a Divergence At Every Step (2017, methodology): “[12] introduce a gradient penalty from the perspective of regret minimization, by setting the regularization function to be a gradient penalty on points around the data manifold, as in Follow The Regularized Leader [3], a standard no-regret algorithm.”
Taking the Human Out of the Loop: A Review of Bayesian Optimization (2016, background): “The book of Cesa-Bianchi [36] is a good reference on the topic of online learning with experts and bandits in adversarial settings.”
Smoothness, Low Noise and Fast Rates (2010, background): “The goal of the player is to have low average objective value 1 n ∑n i=1 `(wi, zi) compared to the best single choice in hind sight [9].”
Same State, Different Task: Continual Reinforcement Learning without Interference (2021, methodology): “We use a modified version of the Exponentially Weighted Average Forecaster algorithm (Cesa-Bianchi and Lugosi 2006), as has been shown to be successful adapting components of RL algorithms (Ball et al. 2020).”
Continual Lifelong Learning in Natural Language Processing: A Survey (2020, background): “+ on-line learning + forward transfer – no backward transfer – no knowledge retention – single task/domain (Bottou, 1999) (Bottou and LeCun, 2004) (Cesa-Bianchi and Lugosi, 2006) (Shalev-Shwartz, 2012) (C.”
Variational Bayesian Reinforcement Learning with Regret Bounds (2018, background): “One way to measure how well an agent balances this trade-off is a quantity called regret, which measures how sub-optimal the rewards the agent has received are so far, relative to the (unknown) optimal policy [12].”
Introduction to Online Convex Optimization (2016, background): “The neighboring book, which served as an inspiration to the current manuscript, and indeed an inspiration to the entire field of learning in games, is the wonderful text of Cesa-Bianchi and Lugosi [29].”
Regret Analysis of Stochastic and Nonstochastic Multi-armed Bandit Problems (2012, neutral): “Now we use the chain rule for Kullback-Leibler divergence —see for example [Cesa-Bianchi and Lugosi, 2006, Section A.2]— iteratively to introduce 36 Adversarial bandits: fundamental results the laws Pti of Y t = (Y1, . . . , Yt).”
Learning Language Games through Interaction (2016, methodology): “While online gradient descent is frequently used, for example in semantic parsing (Zettlemoyer and Collins, 2007; Chen, 2012), we using it in a truly online setting, taking one pass over the data and measuring online accuracy (Cesa-Bianchi and Lugosi, 2006).”
Online Meta-Learning (2019, neutral): “These assumptions are largely standard in online learning, in various settings (Cesa-Bianchi & Lugosi, 2006), except 1.3.”
Stochastic Optimization with Importance Sampling for Regularized Loss Minimization (2014, neutral): “ −hgt−∇f(wt),wt+1 −w∗i ≥ h∇f(wt)+λ∂r(wt+1),w∗ −wt+1i −(1/γ)B ψ(w t+1,wt) −hg t−∇f(wt),wt+1 −w∗i = 1 ηt h∇ψ(wt)−∇ψ(wt+1),w∗ −wt+1i −(1/γ)B ψ(w t+1,wt). Plugging the following equality (Lemma 11.1 from [32]) Bψ(w∗,wt+1) +Bψ(wt+1,wt) −Bψ(w∗,wt) = h∇ψ(wt) −∇ψ(wt+1),w∗ −wt+1i, into the previous inequalit”
On the Sample Complexity of Reinforcement Learning with a Generative Model (2012, methodology): “For this we make use of Bernstein’s concentration inequality (Cesa-Bianchi & Lugosi, 2006, appendix, pg. 361), which bounds the estimation error in terms of the variance of the value function.”
Reinforcement and Imitation Learning via Interactive No-Regret Learning (2014, methodology): “To achieve this, when the policy class Π is ﬁnite, randomized online learning algorithms like weighted majority [9] may be used.”
Fundamental Limits of Online and Distributed Algorithms for Statistical Learning and Estimation (2013, background): “Perhaps the most well-known case is the multi-armed bandits problem with many other variants being developed, such as contextual bandits, combinatorial bandits, and more general models such as partial monitoring [10, 11].”