Explore-only strategy

As explained in Optimization Algorithms: AI techniques for design, planning, and control problems

An explore-only strategy selects actions solely to investigate uncertain alternatives rather than favoring currently estimated rewards.

An explore-only strategy randomly selects an action at every trial without using previous results. In a multi-armed bandit, it gathers information evenly but ignores estimated rewards, creating a clear exploration-only baseline. section 12.1.6

Three slot machines over 90 trials

Suppose three machines have mean payoffs of 8, 10, and 5. Randomly selecting each machine about 30 times produces an expected return of 690, compared with 900 from always selecting machine 2. The resulting regret is 210. section 12.1.6

Figure 12.7 Three slot machines as a noncontextual multi-armed bandit (MAB) problem. The payoff distribution’s mean and standard deviations of these three machines are (8,0.5), (10, 0.7), and (5, 0.45), respectively.

Compare the three arms’ payoff means and standard deviations before implementing the random-selection policy. section 12.1.6

How Optimization Algorithms: AI techniques for design, planning, and control problems covers it

How deep it goes

Explains selection, payoff sampling, regret calculation, and a 90-trial example, but stops short of adaptive exploration.

Examples

Uses three slot machines, Gaussian payoff distributions, 90 trials, NumPy functions, and average-regret comparisons.

What it assumes

Assumes basic reinforcement-learning vocabulary and introduces bandit actions, payoff distributions, exploration, exploitation, and regret.

What you can do after

Apply it via NumPy to simulate random arm selection and measure average regret across trials.

How it connects

Contrasts exploration with exploitation and connects to multi-armed bandits, regret, ε-greedy, greedy selection, and UCB.

liveBookmanning.com

Read Optimization Algorithms: AI techniques for design, planning, and control problems on liveBook Get the book at manning.com

Related concepts

• strategykind of · A strategy is a planned approach for achieving objectives through coordinated choices, priorities, and actions.

This explainer was generated by Manning's knowledge graph from the text of Optimization Algorithms: AI techniques for design, planning, and control problems. The section links point to the parts of the book it draws on.