Multi-armed bandit

Multi-armed bandit
A multi-armed bandit is like a slot machine with multiple levers.

In statistics, particularly in the design of sequential experiments, a multi-armed bandit takes its name from a traditional slot machine (one-armed bandit). Multiple levers are considered in the motivating applications in statistics. When pulled, each lever provides a reward drawn from a distribution associated with that specific lever. The objective of the gambler is to maximize the sum of rewards earned through a sequence of lever pulls.[1][2]

In practice, multi-armed bandits have been used to model the problem of managing research projects in a large organization, like a science foundation or a pharmaceutical company. Given its fixed budget, the problem is to allocate resources among the competing projects, whose properties are only partially known now but may be better understood as time passes.[1][2]

In the early versions of the multi-armed bandit problem, the gambler has no initial knowledge about the levers. The crucial tradeoff the gambler faces at each trial is between "exploitation" of the lever that has the highest expected payoff and "exploration" to get more information about the expected payoffs of the other levers.

The multi-armed bandit is sometimes called a K-armed bandit or N-armed bandit.[citation needed]

Contents

Empirical motivation

The multi-armed bandit problem models an agent that simultaneously attempts to acquire new knowledge and to optimize its decisions based on existing knowledge. There are many practical applications:

In these practical examples, the problem requires balancing reward maximization based on the knowledge already acquired with attempting new actions to further increase knowledge. This is known as the exploitation vs. exploration tradeoff in reinforcement learning.

The model can also be used to control dynamic allocation of resources to different projects, answering the question "which project should I work on" given uncertainty about the difficulty and payoff of each possibility.

Originally considered by Allied scientists in World War II, it proved so intractable that it was proposed the problem be dropped over Germany so that German scientists could also waste their time on it.[3] It was formulated by Herbert Robbins in 1952.

The multi-armed bandit model

The multi-armed bandit (or just bandit for short) can be seen as a set of real distributions B = \{R_1, \dots ,R_K\}, each distribution being associated with the rewards delivered by one of the K levers. Let \mu_1, \dots , \mu_K be the mean values associated with these reward distributions. The gambler iteratively plays one lever per round and observes the associated reward. The objective is to maximize the sum of the collected rewards. The horizon H is the number of rounds that remain to be played. The bandit problem is formally equivalent to a one-state Markov decision process. The regret ρ after T rounds is defined as the difference between the reward sum associated with an optimal strategy and the sum of the collected rewards: \rho = T \mu^* - \sum_{t=1}^T \widehat{r}_t, where μ * is the maximal reward mean, μ * = max kk}, and \widehat{r}_t is the reward at time t. A strategy whose average regret per round ρ / T tends to zero with probability 1 when the number of played rounds tends to infinity is a zero-regret strategy. Intuitively, zero-regret strategies are guaranteed to converge to an optimal strategy, not necessarily unique, if enough rounds are played.

Variations

Another formulation of the multi-armed bandit has each arm representing an independent markov machine. Each time a particular arm is played, the state of that machine advances to a new one, chosen according to the Markov state evolution probabilities. There is a reward depending on the current state of the machine. In a generalisation called the "restless bandit problem", the states of non-played arms can also evolve over time.[4] There has also been discussion of systems where the number of choices (about which arm to play) increases over time.[5]

Computer science researchers have studied multi-armed bandits under worst-case assumptions, obtaining positive[clarification needed] results for finite numbers of trials with both stochastic [6] and nonstochastic[7] arm payoffs.

Common bandit strategies

Many strategies exist which provide an approximate solution to the bandit problem, and can be put into the three broad categories detailed below.

Semi-uniform strategies

Semi-uniform strategies were the earliest (and simplest) strategies discovered to approximately solve the bandit problem. All those strategies have in common a greedy behavior where the best lever (based on previous observations) is always pulled except when a (uniformly) random action is taken.

  • Epsilon-greedy strategy: The best lever is selected for a proportion 1 - \epsilon of the trials, and another lever is randomly selected (with uniform probability) for a proportion \epsilon. A typical parameter value might be \epsilon = 0.1, but this can vary widely depending on circumstances and predilections.
  • Epsilon-first strategy: A pure exploration phase is followed by a pure exploitation phase. For N trials in total, the exploration phase occupies \epsilon N trials and the exploitation phase (1 - \epsilon) N trials. During the exploration phase, a lever is randomly selected (with uniform probability); during the exploitation phase, the best lever is always selected.
  • Epsilon-decreasing strategy: Similar to the epsilon-greedy strategy, except that the value of \epsilon decreases as the experiment progresses, resulting in highly explorative behaviour at the start and highly exploitative behaviour at the finish.
  • Adaptive epsilon-greedy strategy based on value differences (VDBE): Similar to the epsilon-decreasing strategy, except that epsilon is reduced on basis of the learning progress instead of manual tuning (Tokic, 2010). High changes in the value estimates lead to a high epsilon (exploration); low value changes to a low epsilon (exploitation).

Probability matching strategies

Probability matching strategies reflect the idea that the number of pulls for a given lever should match its actual probability of being the optimal lever.

Pricing strategies

Pricing strategies establish a price for each lever. The lever of highest price is always pulled.

See also

Notes

  1. ^ a b c Gittins (1989)
  2. ^ a b c Berry and Fristedt (1985)
  3. ^ Whittle (1979)
  4. ^ Whittle (1988)
  5. ^ Whittle (1981)
  6. ^ P. Auer et al. (2002) "Finite-time analysis of the multiarmed bandit problem." Machine Learning 47:235-256 doi:10.1023/A:1013689704352
  7. ^ P. Auer et al. (2002) "The nonstochastic multiarmed bandit problem." Siam Journal on Computing 32:48-77 doi:10.1137/S0097539701398375

References

External links


Wikimedia Foundation. 2010.

Игры ⚽ Поможем написать реферат

Look at other dictionaries:

  • Bandit (disambiguation) — Bandit as a general term refers to outlaw, other uses include:*Bandit, a Welsh language music television show *A Bandit, the 1913 film starring Fatty Arbuckle *Multi armed bandit, stochastic models of decision theory *Suzuki Bandit Series, Suzuki …   Wikipedia

  • Design of experiments — In general usage, design of experiments (DOE) or experimental design is the design of any information gathering exercises where variation is present, whether under the full control of the experimenter or not. However, in statistics, these terms… …   Wikipedia

  • Search theory — In economics, search theory (or just search) is the study of an individual s optimal strategy when choosing from a series of potential opportunities of random quality, given that delaying choice is costly. Search models illustrate how best to… …   Wikipedia

  • Jack-Pot — Machine à sous Photo d une machine à sous fabriquée en 1960. Une machine à sous est un appareil électronique ou mécanique de jeux de hasard et d argent qui ne demande aucune stratégie ou habileté particulière et dont les lots sont déterminés au… …   Wikipédia en Français

  • Machine a sous — Machine à sous Photo d une machine à sous fabriquée en 1960. Une machine à sous est un appareil électronique ou mécanique de jeux de hasard et d argent qui ne demande aucune stratégie ou habileté particulière et dont les lots sont déterminés au… …   Wikipédia en Français

  • Machine À Sous — Photo d une machine à sous fabriquée en 1960. Une machine à sous est un appareil électronique ou mécanique de jeux de hasard et d argent qui ne demande aucune stratégie ou habileté particulière et dont les lots sont déterminés au lancement du jeu …   Wikipédia en Français

  • Machine à sous — Photo d une machine à sous fabriquée en 1960. Une machine à sous est un appareil électronique ou mécanique de jeux de hasard et d argent qui ne demande aucune stratégie ou habileté particulière. Sommaire 1 …   Wikipédia en Français

  • Machines à sous — Machine à sous Photo d une machine à sous fabriquée en 1960. Une machine à sous est un appareil électronique ou mécanique de jeux de hasard et d argent qui ne demande aucune stratégie ou habileté particulière et dont les lots sont déterminés au… …   Wikipédia en Français

  • Vidéo Poker — Machine à sous Photo d une machine à sous fabriquée en 1960. Une machine à sous est un appareil électronique ou mécanique de jeux de hasard et d argent qui ne demande aucune stratégie ou habileté particulière et dont les lots sont déterminés au… …   Wikipédia en Français

  • Reinforcement learning — Inspired by related psychological theory, in computer science, reinforcement learning is a sub area of machine learning concerned with how an agent ought to take actions in an environment so as to maximize some notion of long term reward .… …   Wikipedia

Share the article and excerpts

Direct link
Do a right-click on the link above
and select “Copy Link”