Dithering for Learning : Computationally Efficient Policies for Dynamic Pricing in High Dimensions
We consider a dynamic pricing and learning problem where a seller prices multiple products and learns from sales data about unknown demand. We propose dithering policies---namely, policies under which prices are probabilistically selected in a neighborhood surrounding the myopic optimal price---that achieve good theoretical performance and are computationally efficient. In particular, by analyzing structural properties of price-dithering, we establish a regret bound of order √T logT. Moreover, we prove under an additional assumption that our policy can be modified to achieve a constant regret bound. With regard to computation, we show that our dithering policies are as efficient as the widely-used myopic policy, yet avoid the problem of incomplete learning that can be encountered under the myopic policy. Dithering, which can be viewed as a "soft" or probabilistic avoidance of incomplete learning, is also shown to be substantially more computationally efficient than existing "discriminative" policies which impose (typically non-convex) hard constraints on the price space and can often be prohibitively time consuming to solve at the optimization stage, particularly in high dimensions
Year of publication: |
2019
|
---|---|
Authors: | Huh, Woonghee Tim |
Other Persons: | Kim, Michael Jong (contributor) ; Lin, Meichun (contributor) |
Publisher: |
[2019]: [S.l.] : SSRN |
Saved in:
freely available
Saved in favorites
Similar items by person
-
Bayesian dithering for learning : asymptotically optimal policies in dynamic pricing
Huh, Woonghee Tim, (2022)
-
Uncertain Search with Knowledge Transfer
Huh, Woonghee Tim, (2023)
-
Multi-period lot-sizing with supplier selection : structural results, complexity and algorithms
Lin, Meichun, (2021)
- More ...