Feature construction is of vital importance in reinforcement learning, as the quality of a value function or policy is largely determined by the corresponding features. The recent successes of deep reinforcement learning (RL) only increase the importance of understanding feature construction. Typical deep RL approaches use a linear output layer, which means that deep RL can be interpreted as a ...

Subset 0x 1x 2x 3x 4x 5x 6x 7x 8x 9x Ax Bx Cx Dx Ex Fx 0x 32 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1x 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2x 0 4 −2 2 −2 2 −4 0 4 0 2 −2 2 −2 0 −4 3x 0 0 −2 6 −2 −2 4 −4 0 0 −2 6 −2 −2 4 −4 4x 0 2 −2 0 0 2 −2 0 0 2 2 4 −4 −2 −2 0 5x 0 2 2 −4 0 10 −6 −4 0 2 −10 0 4 −2 2 4 6x 0 −2 −4 −6 −2 −4 2 0 0 −2 0 −2 −6 −8 2 0 7x 0 2 0 2 −2 8 6 0 −4 6 0 −6 −2 0 −6 −4 8x 0 0 2 6 0 0 −2 −6...

• A submitted manuscript is the author's version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website. • The final author version ...

We derive Sobolev-type inner products with respect to which hat functions on arbitrary triangulations of domains in R are orthogonal. Compared with linear interpolation, the resulting approximation schemes yield superior accuracy at little extra cost.

Space efficient algorithms for the maximum independent set problem for interval graphs and trees are presented in this paper. For a given set of n intervals on a real line, we can compute the maximum independent set in O( 2 s + n log s) time using O(s) extra-space. The lower bound of the time × space product for this problem is Ω(n2− ), where = O(1/ √ log n). We also propose an (1+ 1 ) approxim...