Bandit problems

Pseudopolitical diversion

See clickbait bandits.

Intros

Here’s an intro to all of machine learning through a historical tale about how one particular attempt to teach a machine (not a computer!) to play tic-tac-toe: Rodney Brooks, Machine Learning Explained. Introductions recommended by Bubeck include (Slivkins 2019; Bubeck and Cesa-Bianchi 2012; Lattimore 2020).

Bandits-meet-optimisation

Bubeck again: Kernel-based methods for bandit convex optimization, part 1.

Bandits-meet-evolution

Ben Recht and Roy Frostig, Nesterov’s punctuation equilibrium:

In a landmark new paper by Salimans, Ho, Chen, and Sutskever from OpenAI, (Salimans et al. 2017) the authors show that a particular class of genetic algorithms (called Evolutionary Strategies) gives excellent performance on a variety of reinforcement learning benchmarks. As optimizers, the application of genetic algorithms raises red flags and usually causes us to close browser Windows. But fear not! As we will explain, the particular algorithm deployed happens to be a core method in optimization, and the fact that this method is successful sheds light on the peculiarities of reinforcement learning more than it does about genetic algorithms in general.

Details

🏗

Conceptually, the base model is a one- or many-armed poker machine. You can pop coins in, and each time you do you may pull an arm; you might get rewarded. Each arm of the machine might have different returns; but the only way to find out is to play.

How do you choose optimally which arms to pull and when? How much is to work spending to find the arm with the best return on investment, given that it costs to collect more data?

This can be formalised by minimizing regret and defining some other terms and what you get out is a formalized version of Skinnerian learning that you can easily implement as an algorithm and feel that you have got some satisfyingly optimal properties for.

The thing about whether to choose a new arm when you are on a known good one, or to switch to another one in hope of it being better, this is a symbolic one and it’s called the exploration/exploitation trade-off.

Multi-world testing

A setting by Microsoft: Multi-World Testing (MWT) appears to be an online learning problem that augments its data by re-using the data for offline testing:

… is a toolbox of machine learning technology for principled and efficient experimentation, plausibly applicable to most Microsoft services that interact with customers. In many scenarios, this technology is exponentially more efficient than the traditional A/B testing. The underlying research area, mature and yet very active, is known under many names: “multi-armed bandits”, “contextual bandits”, “associative reinforcement learning”, and “counterfactual evaluation”, among others.

To take an example, suppose one wants to optimize clicks on suggested news stories. To discover what works, one needs to explore over the possible news stories. Further, if the suggested news story can be chosen depending on the visitor’s profile, then one needs to explore over the possible “policies” that map profiles to news stories (and there are exponentially more “policies” than news stories!). Traditional ML fails at this because it does not explore. Whereas MWT allows you to explore continuously, and optimize your decisions using this exploration data.

It has a better introduction here, by John Langford: (cf (Alekh Agarwal et al. 2015).)

Removing the credit assignment problem from reinforcement learning yields the Contextual Bandit setting which we know is generically solvable in the same manner as common supervised learning problems. I know of about a half-dozen real-world successful contextual bandit applications typically requiring the cooperation of engineers and deeply knowledgeable data scientists.

Can we make this dramatically easier? We need a system that explores over appropriate choices with logging of features, actions, probabilities of actions, and outcomes. These must then be fed into an appropriate learning algorithm which trains a policy and then deploys the policy at the point of decision. Naturally, this is what we’ve done and now it can be used by anyone. This drops the barrier to use down to: “Do you have permissions? And do you have a reasonable idea of what a good feature is?”

A key foundational idea is Multiworld Testing: the capability to evaluate large numbers of policies mapping features to action in a manner exponentially more efficient than standard A/B testing. This is used pervasively in the Contextual Bandit literature and you can see it in action for the system we’ve made at Microsoft Research.

Reinforcement learning

See reinforcement learning.

Markov decision problems

See POMDP.

Connection to graphical models

See Levine (2018).

Practicalities

Vowpal Wabbit does contextual bandit learning.:

VW contains a contextual bandit module which allows you to optimize a predictor based on already collected contextual bandit data. In other words, the module does not handle the exploration issue, it assumes it can only use the currently available data previously collected from some “exploration” policy.

Multiple World Testing source code.

Sequential surrogate interactive model optimisation

Not what I think of in reference to bandit problems, but it and many other hyperparameter optimization-type problems have an RL interpretation apparently. That is, you can use RL to learn the hyper parameters of your deep learning model. (This is not the same as deep learning of RL policies.) See Sequential surrogate model optimisation.

Incoming

• Algorithms for Decision Making: Decision making, in the sense of reinforcement learning

References