Multi - Value - Functions : E cient Automatic Action Hierarchies forMultiple Goal
نویسندگان
چکیده
If you have planned to achieve one particular goal in a stochastic delayed rewards problem and then someone asks about a diierent goal what should you do? What if you need to be ready to quickly supply an answer for any possible goal? This paper shows that by using a new kind of automatically generated abstract action hierarchy that with N states, preparing for all of N possible goals can be much much cheaper than N times the work of preparing for one goal. In goal-based Markov Decision Problems, it is usual to generate a policy (x), mapping states to actions, and a value function J(x), mapping states to an estimate of minimum expected cost-to-goal, starting at x. In this paper we will use the terminology that a multi-policy ? (x; y) (for all state-pairs (x; y)) maps a state x to the rst action it should take in order to reach y with expected minimum cost and a multi-value-function J ? (x;y) is a deenition of this minimum cost. Building these objects quickly and with little memory is the main purpose of this paper, but a secondary result is a natural, automatic, way to create a set of parsimonious yet powerful abstract actions for MDPs. The paper concludes with a set of empirical results on increasingly large MDPs.
منابع مشابه
Multi-Value-Functions: E cient Automatic Action Hierarchies for Multiple Goal MDPs
If you have planned to achieve one particular goal in a stochastic delayed rewards problem and then someone asks about a di erent goal what should you do? What if you need to be ready to quickly supply an answer for any possible goal? This paper shows that by using a new kind of automatically generated abstract action hierarchy that with N states, preparing for all of N possible goals can be mu...
متن کاملMulti - Value - Functions : E cient Automatic ActionHierarchies for Multiple Goal MDPs ( Draft of Nov 29 )
متن کامل
Multi-level Association Rule Mining: an Object-oriented Approach Based on Dynamic Hierarchies
Previous studies in data mining have yielded e cient algorithms for discovering association rules. But it is well-known problem that the two controlling measures of support and con dence, when used as the sole de nition of relevant association rules, are too inclusive | interesting rules are included with many uninteresting cases. A typical approach to this problem is to augment the thresholds ...
متن کاملcient Call by value Evaluation Strategy of Primitive Recursive Program Schemes
We consider primitive recursive program schemes with parameters together with the call by value computation rule The schemes are nite systems of functions which are de ned by primitive or structural recursion simultaneous recursion and nesting of function calls is allowed We present a transformation strategy which replaces primitive recursion by iteration The transformation strategy which is fu...
متن کاملAn interactive weighted fuzzy goal programming technique to solve multi-objective reliability optimization problem
This paper presents an application of interactive fuzzy goal programming to the nonlinear multi-objective reliability optimization problem considering system reliability and cost of the system as objective functions. As the decision maker always have an intention to produce highly reliable system with minimum cost, therefore, we introduce the interactive method to design a high productivity sys...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 1999