Advanced Machine Learning
|
|
- Evan Mason
- 5 years ago
- Views:
Transcription
1 Advanced Machine Learning Bandit Problems MEHRYAR MOHRI COURANT INSTITUTE & GOOGLE RESEARCH.
2 Multi-Armed Bandit Problem Problem: which arm of a K-slot machine should a gambler pull to maximize his cumulative reward over a sequence of trials? stochastic setting. adversarial setting. page 2
3 Motivation Clinical trials: potential treatments for a disease to select from, new patient or category at each round (Thompson, 1933). Ads placement: selection of ad to display out of a finite set (which could vary with time though) for each new web page visitor. Adaptive routing: alternative paths for routing packets through a series of tubes or alternative roads for driving from a source to a destination. Games: different moves at each round of a game such as chess, or Go. page 3
4 Key Problem Exploration vs exploitation dilemma (or trade-off): inspect new arms with possibly better rewards. use existing information to select best arm. page 4
5 Outline Stochastic bandits Adversarial bandits page 5
6 Stochastic Model K arms: for each arm i 2 {1,...,K}, reward distribution P i. reward mean µ i. i = µ gap to best:, where µ = max. µ i i2[1,k] µ i page 6
7 Bandit Setting For t =1to T do player selects action (randomized). player receives reward. Equivalent descriptions: I t 2 {1,...,K} X It,t P It on-line learning with partial information ( full). one-state MDPs (Markov Decision Processes). = page 7
8 Objectives Expected regret Pseudo-regret E[R T ]=E By Jensen s inequality, R T apple E[R T ]. " max i2[1,k] R T = max i2[1,k] E = µ T E TX " T X " T X X i,t X i,t X It,t # X T X It,t X T X It,t. # #.. page 8
9 Expected Regret (X i,t µ i ) " If s take values in, then E max i2[1,k] [ r, +r] # TX (X i,t µ ) apple r p 2T log K. The O( p T ) dependency cannot be improved; better guarantees can be achieved for pseudo-regret. page 9
10 Pseudo-Regret Expression in terms of is: KX R T = E[T i (T )] i, where T i (t) denotes the number of times arm i was pulled up to time t, T i (t) = P t. Proof: R T = µ T E " X T =E = i=1 " T X i=1 s=1 1 I s =i X It,t # " T # X =E (µ X It,t) # KX (µ X i,t )1 It =i = " KX X T (µ µ i )E 1 It =i TX KX E[(µ X i,t )] E[1 It =i] i=1 i=1 # = i=1 KX E[T i (T )] i. page 10
11 ε-greedy Strategy At time t, with probability with probability 1 t, select arm i with best emp. mean. t, select random arm. (Auer et al. 2002a) For t =min( 6K 2 t, 1), with = min, 6K for t 2, Pr[I t 6= i ] apple C 2 t for some C>0. E[T i (T )] apple C 2 log T R T apple P i : i>0 thus, and i 2 log T. Logarithmic regret but, requires knowledge of. i : i>0 sub-optimal arms treated similarly (naive search). i C page 11
12 UCB Strategy Optimism in face of uncertainty: at each time compute upper confidence bound (UCB) on the expected reward of each arm. select arm with largest UCB. Idea: wrong arm t 2 [1,T] i cannot be selected for too long. by definition, µ i apple µ apple UCB i. i (Lai and Robbins, 1985; Agrawal 1995; Auer et al. 2002a) µ i pulling often UCB closer to. i 2 [1,K] page 12
13 Note on Concentration Ineqs Let X be a random variable such that for all t 0, log E e t(x E[X]) apple (t), where is a convex function. For Hoeffding s inequality and X 2 [a, b], (t) = t2 (b a) 2. 8 Then, Pr[X E[X] > ] =Pr[e t(x E[X]) >e t ] apple inf t>0 e t E[e t(x E[X]) ] apple inf t>0 e t e (t) = e sup t>0 (t (t)) = e ( )). page 13
14 UCB Strategy Average reward estimate for arm by time : bµ i,t = 1 T i (t) P t s=1 X i,s1 Is =i. Concentration inequality (e.g., Hoeffding s ineq.): Pr[µ i 1 t P t s=1 X i,s > ] apple e t ( ). >0 1 µ i < 1 tx X i,s t t log 1. Thus, for any, with probability at least, s=1 i t page 14
15 (α, ψ)-ucb Strategy >0 Parameter ; -UCB strategy consists of selecting at time t I t 2 argmax i2[1,k] (, ) apple bµ i,t log t T i (t 1). page 15
16 (α, ψ)-ucb Guarantee Theorem: for, the pseudo-regret of -UCB satisfies R T apple for Hoeffding s lemma, -UCB, R T apple >2 X i : i>0 X i : i>0 i 2 ( i (, ) i 2 ) log T + 2 ( ) =2 2 log T (Auer et al. 2002a), page 16
17 Proof Lemma: for any s 0, TX 1 It =i apple s + TX t=s+1 1 It =i 1 Ti (t 1) s. Proof: observe that TX TX 1 It =i = 1 It =i1 Ti (t 1)<s + Now, for t = max {t apple T :1 Ti (t 1)<s 6= 0}, By definition of TX 1 It =i 1 Ti (t 1)<s = t t X TX 1 It =i1 Ti (t 1) s. 1 It =i 1 Ti (t 1)<s., the number of non-zero terms in the sum is at most s. page 17
18 Proof For any and define i,t 1 = 1 log t i t T i (t 1). At time t, if i is selected, then (bµ i,t 1 + i,t 1 ) (bµ i,t + i,t 1) 0,[bµ i,t 1 µ i,t 1 i,t 1 ]+[2 i,t 1 i ]+[µ bµ i,t 1 i,t 1] 0. Thus, at least of one of these three terms is non-negative. Also, if one is non-positive, at least one of the other two is non-negative. page 18
19 Proof To bound the pseudo-regret, we bound observe first that log T log t T i (t 1) s = Thus, apple X T E[T i (T )] = E apple apple s +E apple s + TX t=s+1 1 It =i T X t=s+1 ( i 2 ) 1 It =i 1 Ti (t 1) s ( E[T i (T )]. But, i 2 ) ) i 2 i,t 1 0. Pr[bµ i,t 1 µ i,t 1 i,t 1 0] + Pr[µ bµ i,t 1 i,t 1 0]. page 19
20 Proof Each of the two probability terms can be bounded as follows using the union bound: Pr[µ bµ i,t 1 i,t 1 0] apple apple Pr 9s 2 [1,t]: µ bµ i,s 1 log t s tx 1 apple t = 1 t 1. s=1 0 Final constant of the bound obtained by further simple calculations. page 20
21 Lower Bound Theorem: for any strategy such that E[T i (T )] = o(t ) for any arm i and any >0 for any set of Bernoulli reward distributions, the following holds for all Bernoulli reward distributions: R T X i lim inf T!+1 log T D(µ i k µ ). i : i>0 (Lai and Robbins, 1985) a more general result holds for general distributions. page 21
22 Notes Observe that X i : i>0 i D(µ i k µ ) µ (1 µ ) X i : i>0 1 i, since D(µ i k µ )=µ i log µ i µ +(1 µ i) log 1 µ i 1 µ apple µ i µ i µ µ +(1 µ i ) µ µ i 1 µ = (µ i µ ) 2 µ (1 µ ) = 2 i µ (1 µ ). page 22
23 Outline Stochastic bandits Adversarial bandits page 23
24 Adversarial Model Karms: for each arm i 2 {1,...,K}, no stochastic assumption. rewards in [0, 1]. page 24
25 Bandit Setting For to T do player selects action (randomized). player receives reward. Notes: I t 2 {1,...,K} x It,t rewards x i,t for all arms determined by adversary simultaneously with the selection of an arm by player. adversary oblivious or nonoblivious (or adaptive). T strategies: deterministic, regret of at least 2 for some (bad) sequences, thus must consider randomized. I t page 25
26 Scenarios Oblivious case: adversary rewards selected independently of the player s actions; thus, reward vector at time t only a function of t. Non-oblivious case: adversary rewards at time t function of the player s past actions I 1,...,I t 1. notion of regret problematic: cumulative reward compared to a quantity that depends on the player s actions! (single best action in hindsight function of actions I 1,...,I T played; playing that single best action could have resulted in different rewards.) page 26
27 Objectives Minimize regret ( `i,t =1 x i,t ), expectation or high prob.: R T = max i2[1,k] TX x i,t T X x It,t = TX `It,t min i2[1,k] TX `i,t. Pseudo-regret: " X T R T =E `It,t # min E i2[1,k] " T X `i,t #. R T apple E[R T ] By Jensen s inequality,. page 27
28 Importance Weighting In the bandit setting, the cumulative loss of each arm is not observed, so how should we update the probabilities? Estimates via surrogate loss: è i,t = `i,t 1 It =i, p i,t where p t =(p 1,t,...,p K,t ) is the probability distribution the player uses at time t to draw an arm ( p i,t >0 ). Unbiased estimate: for any i, E [ è KX `i,t i,t] = p j,t 1 j=i = `i,t. I t p t p i,t j=1 page 28
29 EXP3 EXP3(K) 1 p 1 ( 1 K,..., 1 K ) 2 ( L e 1,0,..., L e K,0 ) (0,...,0) 3 for t 1 to T do 4 Sample(I t p t ) 5 Receive(`It,t) 6 for i 1 to K do 7 è i,t `i,t p i,t 1 It =i 8 e Li,t e Li,t 1 + è i,s 9 for i 1 to K do e 10 p L e i,t i,t+1 P K j=1 e L e j,t (Auer et al. 2002b) 11 return p T +1 EXP3 (Exponential weights for Exploration and Exploitation) page 29
30 EXP3 Guarantee Theorem: the pseudo-regret of EXP3 can be bounded as follows: R T apple log K + KT 2. Choosing to minimize the bound gives R T apple p 2KT log K. Proof: similar to that of EG, but we cannot use Hoeffding s inequality since is unbounded. è i,t page 30
31 Proof Potential: t = log P K. Upper bound: t t 1 = log i=1 e e L i,t P K i=1 e L e i,t P K i=1 P N i=1 e L e = log e L e i,t P i,t 1 N i=1 e L e i,t 1 h = log E e è i i,t i p t apple apple 1 e è i,t E e è i,t 1 (log x apple x 1) i pt E i pt è i,t è 2 i,t (e x apple 1 x + x2 2 ) = E i pt [ è i,t]+ 2 2 E i p t = `It,t l 2 I t,t p It,t apple apple l 2 i,t 1 It =i p 2 i,t `It,t p It,t. page 31
32 Proof Upper bound: summing up the inequalities yields apple X T E[ T 0] apple E It p t `It,t + E It p t apple T X 2 2p It,t apple X T = E `It,t + 2 KT 2. Lower bound: for all j 2 [1,K], E[ T 0] = E It p t Comparison: 8j 2 [1,K], apple apple X K log i=1 e e L i,t log K E [ L e j,t ] log K = E [L j,t ] log K. It p t It p t apple X T E `It,t E[L j,t ] apple log K KT ) R T apple log K + KT 2. page 32
33 Notes When T is not known: standard doubling trick. q log K t = Kt or, use, then R T apple 2 p KT log K. High probability bounds: importance weighting problem: unbounded second moment (see (Cortes, Mansour, MM, 2010)), (Auer et al., 2002b): mixing probability with a uniform distribution to ensure a lower bound on p i,t ; but not sufficient for high probability bound. solution: biased estimate p i,t with >0 a parameter to tune. è i,t = `i,t1 It =i+ E i pt [ è 2 i,t ]= `2 I t,t p It,t. page 33
34 Lower Bound (Bubek and Cesa-Bianchi, 2012) Sufficient lower bound in a stochastic setting for the pseudo-regret (and therefore for the expected regret). Theorem: for any T 1 and any player strategy, there exists a distribution of losses in {0, 1} for which 1 p R T KT. 20 page 34
35 Notes Bound of EXP3 matching lower bound modulo Log term. Log-free bound: p i,t+1 = (C t Li,t e ) where C is a constant P t K ensuring i=1 p i,t+1 =1 and increasing, convex, twice differentiable over (Audibert and Bubeck, 2010). R EXP3 coincides with (x) =e x. formulation as mirror descent. only in oblivious case. log-free bound with (x) =( x) q and q =2. page 35
36 References R. Agrawal. Sample mean based index policies with O(log n) regret for the multi-armed bandit problem. Advances in Applied Mathematics, vol. 27, pp , Jean-Yves Audibert and Sebastian Bubeck. Regret bounds and minimax policies under partial monitoring. Journal of Machine Learning Research, vol. 11, pp , Peter Auer, Nicolò Cesa-Bianchi, and Paul Fischer. Finite-time analysis of the multi- armed bandit problem, Machine Learning Journal, vol. 47, no. 2 3, pp , 2002a. Peter Auer, Nicolò Cesa-Bianchi, Yoav Freund, and Robert Schapire. The nonstochastic multi-armed bandit problem. SIAM Journal on Computing, vol. 32, no. 1, pp , 2002b. Sébastian Bubeck, Nicolò Cesa-Bianchi. Regret analysis of stochastic and nonstochastic multi-armed bandit problems. Foundations and Trends in Machine Learning 5, 1-122, page 36
37 References Nicolò Cesa-Bianchi and Gábor Lugosi. Prediction, learning, and games. Cambridge University Press, Corinna Cortes, Yishay Mansour, and Mehryar Mohri. Learning bounds for importance weighting. In NIPS, T.L. Lai and H. Robbins. Asymptotically efficient adaptive allocation rules. Advances in Applied Mathematics, vol. 6, pp. 4 22, Gilles Stoltz. Incomplete information and internal regret in prediction of individual sequences. Ph.D. thesis, Universite Paris-Sud, R. Tyrrell Rockafellar. Convex Analysis. Princeton University Press, W. Thompson. On the likelihood that one unknown probability exceeds another in view of the evidence of two samples. Bulletin of the American Mathematics Society, vol. 25, pp , page 37
Regret Analysis of Stochastic and Nonstochastic Multi-armed Bandit Problems, Part I. Sébastien Bubeck Theory Group
Regret Analysis of Stochastic and Nonstochastic Multi-armed Bandit Problems, Part I Sébastien Bubeck Theory Group i.i.d. multi-armed bandit, Robbins [1952] i.i.d. multi-armed bandit, Robbins [1952] Known
More informationBandit models: a tutorial
Gdt COS, December 3rd, 2015 Multi-Armed Bandit model: general setting K arms: for a {1,..., K}, (X a,t ) t N is a stochastic process. (unknown distributions) Bandit game: a each round t, an agent chooses
More informationMulti-armed bandit models: a tutorial
Multi-armed bandit models: a tutorial CERMICS seminar, March 30th, 2016 Multi-Armed Bandit model: general setting K arms: for a {1,..., K}, (X a,t ) t N is a stochastic process. (unknown distributions)
More informationThe Multi-Armed Bandit Problem
Università degli Studi di Milano The bandit problem [Robbins, 1952]... K slot machines Rewards X i,1, X i,2,... of machine i are i.i.d. [0, 1]-valued random variables An allocation policy prescribes which
More informationStat 260/CS Learning in Sequential Decision Problems. Peter Bartlett
Stat 260/CS 294-102. Learning in Sequential Decision Problems. Peter Bartlett 1. Adversarial bandits Definition: sequential game. Lower bounds on regret from the stochastic case. Exp3: exponential weights
More informationOnline Learning with Feedback Graphs
Online Learning with Feedback Graphs Claudio Gentile INRIA and Google NY clagentile@gmailcom NYC March 6th, 2018 1 Content of this lecture Regret analysis of sequential prediction problems lying between
More informationLecture 19: UCB Algorithm and Adversarial Bandit Problem. Announcements Review on stochastic multi-armed bandit problem
Lecture 9: UCB Algorithm and Adversarial Bandit Problem EECS598: Prediction and Learning: It s Only a Game Fall 03 Lecture 9: UCB Algorithm and Adversarial Bandit Problem Prof. Jacob Abernethy Scribe:
More informationThe Multi-Armed Bandit Problem
The Multi-Armed Bandit Problem Electrical and Computer Engineering December 7, 2013 Outline 1 2 Mathematical 3 Algorithm Upper Confidence Bound Algorithm A/B Testing Exploration vs. Exploitation Scientist
More informationEASINESS IN BANDITS. Gergely Neu. Pompeu Fabra University
EASINESS IN BANDITS Gergely Neu Pompeu Fabra University EASINESS IN BANDITS Gergely Neu Pompeu Fabra University THE BANDIT PROBLEM Play for T rounds attempting to maximize rewards THE BANDIT PROBLEM Play
More informationIntroduction to Bandit Algorithms. Introduction to Bandit Algorithms
Stochastic K-Arm Bandit Problem Formulation Consider K arms (actions) each correspond to an unknown distribution {ν k } K k=1 with values bounded in [0, 1]. At each time t, the agent pulls an arm I t {1,...,
More informationGambling in a rigged casino: The adversarial multi-armed bandit problem
Gambling in a rigged casino: The adversarial multi-armed bandit problem Peter Auer Institute for Theoretical Computer Science University of Technology Graz A-8010 Graz (Austria) pauer@igi.tu-graz.ac.at
More informationThe Multi-Arm Bandit Framework
The Multi-Arm Bandit Framework A. LAZARIC (SequeL Team @INRIA-Lille) ENS Cachan - Master 2 MVA SequeL INRIA Lille MVA-RL Course In This Lecture A. LAZARIC Reinforcement Learning Algorithms Oct 29th, 2013-2/94
More informationAdvanced Topics in Machine Learning and Algorithmic Game Theory Fall semester, 2011/12
Advanced Topics in Machine Learning and Algorithmic Game Theory Fall semester, 2011/12 Lecture 4: Multiarmed Bandit in the Adversarial Model Lecturer: Yishay Mansour Scribe: Shai Vardi 4.1 Lecture Overview
More informationAn Estimation Based Allocation Rule with Super-linear Regret and Finite Lock-on Time for Time-dependent Multi-armed Bandit Processes
An Estimation Based Allocation Rule with Super-linear Regret and Finite Lock-on Time for Time-dependent Multi-armed Bandit Processes Prokopis C. Prokopiou, Peter E. Caines, and Aditya Mahajan McGill University
More informationBandit Algorithms. Zhifeng Wang ... Department of Statistics Florida State University
Bandit Algorithms Zhifeng Wang Department of Statistics Florida State University Outline Multi-Armed Bandits (MAB) Exploration-First Epsilon-Greedy Softmax UCB Thompson Sampling Adversarial Bandits Exp3
More informationAdvanced Machine Learning
Advanced Machine Learning Learning and Games MEHRYAR MOHRI MOHRI@ COURANT INSTITUTE & GOOGLE RESEARCH. Outline Normal form games Nash equilibrium von Neumann s minimax theorem Correlated equilibrium Internal
More informationCOS 402 Machine Learning and Artificial Intelligence Fall Lecture 22. Exploration & Exploitation in Reinforcement Learning: MAB, UCB, Exp3
COS 402 Machine Learning and Artificial Intelligence Fall 2016 Lecture 22 Exploration & Exploitation in Reinforcement Learning: MAB, UCB, Exp3 How to balance exploration and exploitation in reinforcement
More informationThe information complexity of sequential resource allocation
The information complexity of sequential resource allocation Emilie Kaufmann, joint work with Olivier Cappé, Aurélien Garivier and Shivaram Kalyanakrishan SMILE Seminar, ENS, June 8th, 205 Sequential allocation
More informationBandit Algorithms. Tor Lattimore & Csaba Szepesvári
Bandit Algorithms Tor Lattimore & Csaba Szepesvári Bandits Time 1 2 3 4 5 6 7 8 9 10 11 12 Left arm $1 $0 $1 $1 $0 Right arm $1 $0 Five rounds to go. Which arm would you play next? Overview What are bandits,
More informationStat 260/CS Learning in Sequential Decision Problems. Peter Bartlett
Stat 260/CS 294-102. Learning in Sequential Decision Problems. Peter Bartlett 1. Multi-armed bandit algorithms. Concentration inequalities. P(X ǫ) exp( ψ (ǫ))). Cumulant generating function bounds. Hoeffding
More informationOn the Complexity of Best Arm Identification in Multi-Armed Bandit Models
On the Complexity of Best Arm Identification in Multi-Armed Bandit Models Aurélien Garivier Institut de Mathématiques de Toulouse Information Theory, Learning and Big Data Simons Institute, Berkeley, March
More informationTwo optimization problems in a stochastic bandit model
Two optimization problems in a stochastic bandit model Emilie Kaufmann joint work with Olivier Cappé, Aurélien Garivier and Shivaram Kalyanakrishnan Journées MAS 204, Toulouse Outline From stochastic optimization
More informationExponential Weights on the Hypercube in Polynomial Time
European Workshop on Reinforcement Learning 14 (2018) October 2018, Lille, France. Exponential Weights on the Hypercube in Polynomial Time College of Information and Computer Sciences University of Massachusetts
More informationYevgeny Seldin. University of Copenhagen
Yevgeny Seldin University of Copenhagen Classical (Batch) Machine Learning Collect Data Data Assumption The samples are independent identically distributed (i.i.d.) Machine Learning Prediction rule New
More informationAlireza Shafaei. Machine Learning Reading Group The University of British Columbia Summer 2017
s s Machine Learning Reading Group The University of British Columbia Summer 2017 (OCO) Convex 1/29 Outline (OCO) Convex Stochastic Bernoulli s (OCO) Convex 2/29 At each iteration t, the player chooses
More informationOnline Learning and Sequential Decision Making
Online Learning and Sequential Decision Making Emilie Kaufmann CNRS & CRIStAL, Inria SequeL, emilie.kaufmann@univ-lille.fr Research School, ENS Lyon, Novembre 12-13th 2018 Emilie Kaufmann Sequential Decision
More informationTHE first formalization of the multi-armed bandit problem
EDIC RESEARCH PROPOSAL 1 Multi-armed Bandits in a Network Farnood Salehi I&C, EPFL Abstract The multi-armed bandit problem is a sequential decision problem in which we have several options (arms). We can
More informationRevisiting the Exploration-Exploitation Tradeoff in Bandit Models
Revisiting the Exploration-Exploitation Tradeoff in Bandit Models joint work with Aurélien Garivier (IMT, Toulouse) and Tor Lattimore (University of Alberta) Workshop on Optimization and Decision-Making
More informationMulti-Armed Bandit Formulations for Identification and Control
Multi-Armed Bandit Formulations for Identification and Control Cristian R. Rojas Joint work with Matías I. Müller and Alexandre Proutiere KTH Royal Institute of Technology, Sweden ERNSI, September 24-27,
More informationReinforcement Learning
Reinforcement Learning Lecture 5: Bandit optimisation Alexandre Proutiere, Sadegh Talebi, Jungseul Ok KTH, The Royal Institute of Technology Objectives of this lecture Introduce bandit optimisation: the
More informationAnytime optimal algorithms in stochastic multi-armed bandits
Rémy Degenne LPMA, Université Paris Diderot Vianney Perchet CREST, ENSAE REMYDEGENNE@MATHUNIV-PARIS-DIDEROTFR VIANNEYPERCHET@NORMALESUPORG Abstract We introduce an anytime algorithm for stochastic multi-armed
More informationLecture 4: Lower Bounds (ending); Thompson Sampling
CMSC 858G: Bandits, Experts and Games 09/12/16 Lecture 4: Lower Bounds (ending); Thompson Sampling Instructor: Alex Slivkins Scribed by: Guowei Sun,Cheng Jie 1 Lower bounds on regret (ending) Recap from
More informationRegret Analysis of Stochastic and Nonstochastic Multi-armed Bandit Problems. Sébastien Bubeck Theory Group
Regret Analysis of Stochastic and Nonstochastic Multi-armed Bandit Problems Sébastien Bubeck Theory Group Part 1: i.i.d., adversarial, and Bayesian bandit models i.i.d. multi-armed bandit, Robbins [1952]
More informationCsaba Szepesvári 1. University of Alberta. Machine Learning Summer School, Ile de Re, France, 2008
LEARNING THEORY OF OPTIMAL DECISION MAKING PART I: ON-LINE LEARNING IN STOCHASTIC ENVIRONMENTS Csaba Szepesvári 1 1 Department of Computing Science University of Alberta Machine Learning Summer School,
More informationEvaluation of multi armed bandit algorithms and empirical algorithm
Acta Technica 62, No. 2B/2017, 639 656 c 2017 Institute of Thermomechanics CAS, v.v.i. Evaluation of multi armed bandit algorithms and empirical algorithm Zhang Hong 2,3, Cao Xiushan 1, Pu Qiumei 1,4 Abstract.
More informationFoundations of Machine Learning On-Line Learning. Mehryar Mohri Courant Institute and Google Research
Foundations of Machine Learning On-Line Learning Mehryar Mohri Courant Institute and Google Research mohri@cims.nyu.edu Motivation PAC learning: distribution fixed over time (training and test). IID assumption.
More informationStratégies bayésiennes et fréquentistes dans un modèle de bandit
Stratégies bayésiennes et fréquentistes dans un modèle de bandit thèse effectuée à Telecom ParisTech, co-dirigée par Olivier Cappé, Aurélien Garivier et Rémi Munos Journées MAS, Grenoble, 30 août 2016
More informationMultiple Identifications in Multi-Armed Bandits
Multiple Identifications in Multi-Armed Bandits arxiv:05.38v [cs.lg] 4 May 0 Sébastien Bubeck Department of Operations Research and Financial Engineering, Princeton University sbubeck@princeton.edu Tengyao
More informationExploration. 2015/10/12 John Schulman
Exploration 2015/10/12 John Schulman What is the exploration problem? Given a long-lived agent (or long-running learning algorithm), how to balance exploration and exploitation to maximize long-term rewards
More informationNew Algorithms for Contextual Bandits
New Algorithms for Contextual Bandits Lev Reyzin Georgia Institute of Technology Work done at Yahoo! 1 S A. Beygelzimer, J. Langford, L. Li, L. Reyzin, R.E. Schapire Contextual Bandit Algorithms with Supervised
More informationStochastic bandits: Explore-First and UCB
CSE599s, Spring 2014, Online Learning Lecture 15-2/19/2014 Stochastic bandits: Explore-First and UCB Lecturer: Brendan McMahan or Ofer Dekel Scribe: Javad Hosseini In this lecture, we like to answer this
More informationAdvanced Machine Learning
Advanced Machine Learning Online Convex Optimization MEHRYAR MOHRI MOHRI@ COURANT INSTITUTE & GOOGLE RESEARCH. Outline Online projected sub-gradient descent. Exponentiated Gradient (EG). Mirror descent.
More informationLearning to play K-armed bandit problems
Learning to play K-armed bandit problems Francis Maes 1, Louis Wehenkel 1 and Damien Ernst 1 1 University of Liège Dept. of Electrical Engineering and Computer Science Institut Montefiore, B28, B-4000,
More informationTwo generic principles in modern bandits: the optimistic principle and Thompson sampling
Two generic principles in modern bandits: the optimistic principle and Thompson sampling Rémi Munos INRIA Lille, France CSML Lunch Seminars, September 12, 2014 Outline Two principles: The optimistic principle
More informationCombinatorial Multi-Armed Bandit: General Framework, Results and Applications
Combinatorial Multi-Armed Bandit: General Framework, Results and Applications Wei Chen Microsoft Research Asia, Beijing, China Yajun Wang Microsoft Research Asia, Beijing, China Yang Yuan Computer Science
More informationMulti-Armed Bandits. Credit: David Silver. Google DeepMind. Presenter: Tianlu Wang
Multi-Armed Bandits Credit: David Silver Google DeepMind Presenter: Tianlu Wang Credit: David Silver (DeepMind) Multi-Armed Bandits Presenter: Tianlu Wang 1 / 27 Outline 1 Introduction Exploration vs.
More informationLearning, Games, and Networks
Learning, Games, and Networks Abhishek Sinha Laboratory for Information and Decision Systems MIT ML Talk Series @CNRG December 12, 2016 1 / 44 Outline 1 Prediction With Experts Advice 2 Application to
More informationLecture 3: Lower Bounds for Bandit Algorithms
CMSC 858G: Bandits, Experts and Games 09/19/16 Lecture 3: Lower Bounds for Bandit Algorithms Instructor: Alex Slivkins Scribed by: Soham De & Karthik A Sankararaman 1 Lower Bounds In this lecture (and
More informationarxiv: v2 [cs.lg] 3 Nov 2012
arxiv:1204.5721v2 [cs.lg] 3 Nov 2012 Regret Analysis of Stochastic and Nonstochastic Multi-armed Bandit Problems Sébastien Bubeck 1 and Nicolò Cesa-Bianchi 2 1 Department of Operations Research and Financial
More informationBandits for Online Optimization
Bandits for Online Optimization Nicolò Cesa-Bianchi Università degli Studi di Milano N. Cesa-Bianchi (UNIMI) Bandits for Online Optimization 1 / 16 The multiarmed bandit problem... K slot machines Each
More informationLearning Algorithms for Minimizing Queue Length Regret
Learning Algorithms for Minimizing Queue Length Regret Thomas Stahlbuhk Massachusetts Institute of Technology Cambridge, MA Brooke Shrader MIT Lincoln Laboratory Lexington, MA Eytan Modiano Massachusetts
More informationAdvanced Machine Learning
Advanced Machine Learning Follow-he-Perturbed Leader MEHRYAR MOHRI MOHRI@ COURAN INSIUE & GOOGLE RESEARCH. General Ideas Linear loss: decomposition as a sum along substructures. sum of edge losses in a
More informationThe information complexity of best-arm identification
The information complexity of best-arm identification Emilie Kaufmann, joint work with Olivier Cappé and Aurélien Garivier MAB workshop, Lancaster, January th, 206 Context: the multi-armed bandit model
More informationBandits : optimality in exponential families
Bandits : optimality in exponential families Odalric-Ambrym Maillard IHES, January 2016 Odalric-Ambrym Maillard Bandits 1 / 40 Introduction 1 Stochastic multi-armed bandits 2 Boundary crossing probabilities
More informationExplore no more: Improved high-probability regret bounds for non-stochastic bandits
Explore no more: Improved high-probability regret bounds for non-stochastic bandits Gergely Neu SequeL team INRIA Lille Nord Europe gergely.neu@gmail.com Abstract This work addresses the problem of regret
More informationStochastic Contextual Bandits with Known. Reward Functions
Stochastic Contextual Bandits with nown 1 Reward Functions Pranav Sakulkar and Bhaskar rishnamachari Ming Hsieh Department of Electrical Engineering Viterbi School of Engineering University of Southern
More informationInformational Confidence Bounds for Self-Normalized Averages and Applications
Informational Confidence Bounds for Self-Normalized Averages and Applications Aurélien Garivier Institut de Mathématiques de Toulouse - Université Paul Sabatier Thursday, September 12th 2013 Context Tree
More informationBandits and Exploration: How do we (optimally) gather information? Sham M. Kakade
Bandits and Exploration: How do we (optimally) gather information? Sham M. Kakade Machine Learning for Big Data CSE547/STAT548 University of Washington S. M. Kakade (UW) Optimization for Big data 1 / 22
More informationImproved Algorithms for Linear Stochastic Bandits
Improved Algorithms for Linear Stochastic Bandits Yasin Abbasi-Yadkori abbasiya@ualberta.ca Dept. of Computing Science University of Alberta Dávid Pál dpal@google.com Dept. of Computing Science University
More informationBayesian and Frequentist Methods in Bandit Models
Bayesian and Frequentist Methods in Bandit Models Emilie Kaufmann, Telecom ParisTech Bayes In Paris, ENSAE, October 24th, 2013 Emilie Kaufmann (Telecom ParisTech) Bayesian and Frequentist Bandits BIP,
More informationRegional Multi-Armed Bandits
School of Information Science and Technology University of Science and Technology of China {wzy43, zrd7}@mail.ustc.edu.cn, congshen@ustc.edu.cn Abstract We consider a variant of the classic multiarmed
More informationOn the Complexity of Best Arm Identification with Fixed Confidence
On the Complexity of Best Arm Identification with Fixed Confidence Discrete Optimization with Noise Aurélien Garivier, joint work with Emilie Kaufmann CNRS, CRIStAL) to be presented at COLT 16, New York
More informationBest Arm Identification in Multi-Armed Bandits
Best Arm Identification in Multi-Armed Bandits Jean-Yves Audibert Imagine, Université Paris Est & Willow, CNRS/ENS/INRIA, Paris, France audibert@certisenpcfr Sébastien Bubeck, Rémi Munos SequeL Project,
More informationLogarithmic Online Regret Bounds for Undiscounted Reinforcement Learning
Logarithmic Online Regret Bounds for Undiscounted Reinforcement Learning Peter Auer Ronald Ortner University of Leoben, Franz-Josef-Strasse 18, 8700 Leoben, Austria auer,rortner}@unileoben.ac.at Abstract
More informationMistake Bounds on Noise-Free Multi-Armed Bandit Game
Mistake Bounds on Noise-Free Multi-Armed Bandit Game Atsuyoshi Nakamura a,, David P. Helmbold b, Manfred K. Warmuth b a Hokkaido University, Kita 14, Nishi 9, Kita-ku, Sapporo 060-0814, Japan b University
More informationMulti armed bandit problem: some insights
Multi armed bandit problem: some insights July 4, 20 Introduction Multi Armed Bandit problems have been widely studied in the context of sequential analysis. The application areas include clinical trials,
More informationGrundlagen der Künstlichen Intelligenz
Grundlagen der Künstlichen Intelligenz Uncertainty & Probabilities & Bandits Daniel Hennes 16.11.2017 (WS 2017/18) University Stuttgart - IPVS - Machine Learning & Robotics 1 Today Uncertainty Probability
More informationConditional Swap Regret and Conditional Correlated Equilibrium
Conditional Swap Regret and Conditional Correlated quilibrium Mehryar Mohri Courant Institute and Google 251 Mercer Street New York, NY 10012 mohri@cims.nyu.edu Scott Yang Courant Institute 251 Mercer
More informationSubsampling, Concentration and Multi-armed bandits
Subsampling, Concentration and Multi-armed bandits Odalric-Ambrym Maillard, R. Bardenet, S. Mannor, A. Baransi, N. Galichet, J. Pineau, A. Durand Toulouse, November 09, 2015 O-A. Maillard Subsampling and
More informationPure Exploration in Finitely Armed and Continuous Armed Bandits
Pure Exploration in Finitely Armed and Continuous Armed Bandits Sébastien Bubeck INRIA Lille Nord Europe, SequeL project, 40 avenue Halley, 59650 Villeneuve d Ascq, France Rémi Munos INRIA Lille Nord Europe,
More informationMulti-armed Bandits in the Presence of Side Observations in Social Networks
52nd IEEE Conference on Decision and Control December 0-3, 203. Florence, Italy Multi-armed Bandits in the Presence of Side Observations in Social Networks Swapna Buccapatnam, Atilla Eryilmaz, and Ness
More informationTsinghua Machine Learning Guest Lecture, June 9,
Tsinghua Machine Learning Guest Lecture, June 9, 2015 1 Lecture Outline Introduction: motivations and definitions for online learning Multi-armed bandit: canonical example of online learning Combinatorial
More informationTheory and Applications of A Repeated Game Playing Algorithm. Rob Schapire Princeton University [currently visiting Yahoo!
Theory and Applications of A Repeated Game Playing Algorithm Rob Schapire Princeton University [currently visiting Yahoo! Research] Learning Is (Often) Just a Game some learning problems: learn from training
More informationExplore no more: Improved high-probability regret bounds for non-stochastic bandits
Explore no more: Improved high-probability regret bounds for non-stochastic bandits Gergely Neu SequeL team INRIA Lille Nord Europe gergely.neu@gmail.com Abstract This work addresses the problem of regret
More informationOn Bayesian bandit algorithms
On Bayesian bandit algorithms Emilie Kaufmann joint work with Olivier Cappé, Aurélien Garivier, Nathaniel Korda and Rémi Munos July 1st, 2012 Emilie Kaufmann (Telecom ParisTech) On Bayesian bandit algorithms
More informationToward a Classification of Finite Partial-Monitoring Games
Toward a Classification of Finite Partial-Monitoring Games Gábor Bartók ( *student* ), Dávid Pál, and Csaba Szepesvári Department of Computing Science, University of Alberta, Canada {bartok,dpal,szepesva}@cs.ualberta.ca
More informationAdvanced Machine Learning
Advanced Machine Learning Learning with Large Expert Spaces MEHRYAR MOHRI MOHRI@ COURANT INSTITUTE & GOOGLE RESEARCH. Problem Learning guarantees: R T = O( p T log N). informative even for N very large.
More informationThe No-Regret Framework for Online Learning
The No-Regret Framework for Online Learning A Tutorial Introduction Nahum Shimkin Technion Israel Institute of Technology Haifa, Israel Stochastic Processes in Engineering IIT Mumbai, March 2013 N. Shimkin,
More informationDownloaded 11/11/14 to Redistribution subject to SIAM license or copyright; see
SIAM J. COMPUT. Vol. 32, No. 1, pp. 48 77 c 2002 Society for Industrial and Applied Mathematics THE NONSTOCHASTIC MULTIARMED BANDIT PROBLEM PETER AUER, NICOLÒ CESA-BIANCHI, YOAV FREUND, AND ROBERT E. SCHAPIRE
More informationOnline Learning with Gaussian Payoffs and Side Observations
Online Learning with Gaussian Payoffs and Side Observations Yifan Wu 1 András György 2 Csaba Szepesvári 1 1 Department of Computing Science University of Alberta 2 Department of Electrical and Electronic
More informationExploration and exploitation of scratch games
Mach Learn (2013) 92:377 401 DOI 10.1007/s10994-013-5359-2 Exploration and exploitation of scratch games Raphaël Féraud Tanguy Urvoy Received: 10 January 2013 / Accepted: 12 April 2013 / Published online:
More informationChapter 2 Stochastic Multi-armed Bandit
Chapter 2 Stochastic Multi-armed Bandit Abstract In this chapter, we present the formulation, theoretical bound, and algorithms for the stochastic MAB problem. Several important variants of stochastic
More informationPure Exploration Stochastic Multi-armed Bandits
C&A Workshop 2016, Hangzhou Pure Exploration Stochastic Multi-armed Bandits Jian Li Institute for Interdisciplinary Information Sciences Tsinghua University Outline Introduction 2 Arms Best Arm Identification
More information1 MDP Value Iteration Algorithm
CS 0. - Active Learning Problem Set Handed out: 4 Jan 009 Due: 9 Jan 009 MDP Value Iteration Algorithm. Implement the value iteration algorithm given in the lecture. That is, solve Bellman s equation using
More informationLearning Methods for Online Prediction Problems. Peter Bartlett Statistics and EECS UC Berkeley
Learning Methods for Online Prediction Problems Peter Bartlett Statistics and EECS UC Berkeley Course Synopsis A finite comparison class: A = {1,..., m}. Converting online to batch. Online convex optimization.
More informationA Drifting-Games Analysis for Online Learning and Applications to Boosting
A Drifting-Games Analysis for Online Learning and Applications to Boosting Haipeng Luo Department of Computer Science Princeton University Princeton, NJ 08540 haipengl@cs.princeton.edu Robert E. Schapire
More informationRegret lower bounds and extended Upper Confidence Bounds policies in stochastic multi-armed bandit problem
Journal of Machine Learning Research 1 01) 1-48 Submitted 4/00; Published 10/00 Regret lower bounds and extended Upper Confidence Bounds policies in stochastic multi-armed bandit problem Antoine Salomon
More informationFoundations of Machine Learning
Introduction to ML Mehryar Mohri Courant Institute and Google Research mohri@cims.nyu.edu page 1 Logistics Prerequisites: basics in linear algebra, probability, and analysis of algorithms. Workload: about
More informationReinforcement Learning
Reinforcement Learning Markov decision process & Dynamic programming Evaluative feedback, value function, Bellman equation, optimality, Markov property, Markov decision process, dynamic programming, value
More informationApplications of on-line prediction. in telecommunication problems
Applications of on-line prediction in telecommunication problems Gábor Lugosi Pompeu Fabra University, Barcelona based on joint work with András György and Tamás Linder 1 Outline On-line prediction; Some
More informationStochastic Multi-Armed-Bandit Problem with Non-stationary Rewards
000 001 002 003 004 005 006 007 008 009 010 011 012 013 014 015 016 017 018 019 020 021 022 023 024 025 026 027 028 029 030 031 032 033 034 035 036 037 038 039 040 041 042 043 044 045 046 047 048 049 050
More informationBandit View on Continuous Stochastic Optimization
Bandit View on Continuous Stochastic Optimization Sébastien Bubeck 1 joint work with Rémi Munos 1 & Gilles Stoltz 2 & Csaba Szepesvari 3 1 INRIA Lille, SequeL team 2 CNRS/ENS/HEC 3 University of Alberta
More informationAnalysis of Thompson Sampling for the multi-armed bandit problem
Analysis of Thompson Sampling for the multi-armed bandit problem Shipra Agrawal Microsoft Research India shipra@microsoft.com avin Goyal Microsoft Research India navingo@microsoft.com Abstract We show
More informationA Second-order Bound with Excess Losses
A Second-order Bound with Excess Losses Pierre Gaillard 12 Gilles Stoltz 2 Tim van Erven 3 1 EDF R&D, Clamart, France 2 GREGHEC: HEC Paris CNRS, Jouy-en-Josas, France 3 Leiden University, the Netherlands
More informationIntroduction to Reinforcement Learning Part 3: Exploration for decision making, Application to games, optimization, and planning
Introduction to Reinforcement Learning Part 3: Exploration for decision making, Application to games, optimization, and planning Rémi Munos SequeL project: Sequential Learning http://researchers.lille.inria.fr/
More informationAdministration. CSCI567 Machine Learning (Fall 2018) Outline. Outline. HW5 is available, due on 11/18. Practice final will also be available soon.
Administration CSCI567 Machine Learning Fall 2018 Prof. Haipeng Luo U of Southern California Nov 7, 2018 HW5 is available, due on 11/18. Practice final will also be available soon. Remaining weeks: 11/14,
More informationDeviations of stochastic bandit regret
Deviations of stochastic bandit regret Antoine Salomon 1 and Jean-Yves Audibert 1,2 1 Imagine École des Ponts ParisTech Université Paris Est salomona@imagine.enpc.fr audibert@imagine.enpc.fr 2 Sierra,
More informationTutorial: PART 2. Online Convex Optimization, A Game- Theoretic Approach to Learning
Tutorial: PART 2 Online Convex Optimization, A Game- Theoretic Approach to Learning Elad Hazan Princeton University Satyen Kale Yahoo Research Exploiting curvature: logarithmic regret Logarithmic regret
More informationOptimistic Bayesian Sampling in Contextual-Bandit Problems
Journal of Machine Learning Research volume (2012) 2069-2106 Submitted 7/11; Revised 5/12; Published 6/12 Optimistic Bayesian Sampling in Contextual-Bandit Problems Benedict C. May School of Mathematics
More informationOn the Complexity of Best Arm Identification with Fixed Confidence
On the Complexity of Best Arm Identification with Fixed Confidence Discrete Optimization with Noise Aurélien Garivier, Emilie Kaufmann COLT, June 23 th 2016, New York Institut de Mathématiques de Toulouse
More information