Sampling device and sampling method
AssigneeInter-University Research Institute Corporation Research Organization of Information and Systems
InventorsIchiro Hasuo, Yuichiro Oyabu, Clovis Eberhart, Kohei Suenaga, Kenta Cho, Shinya Katsumata
Life of this patent
Estimating the term…
Abstract
A sampling device includes a memory and a processor configured to execute selecting a control flow corresponding to each branch in a probabilistic program; optimizing the program by applying predetermined transformation rules to the program of the selected control flow by back propagation; and generating samples according to the optimized program.
Claims8
A sampling device comprising:
- a memory;
- and
- a processor configured to:
- transform instructions of a serial program of a probabilistic program by using an internal state in reverse order of the serial program, the internal state being determined based on an instruction to be transformed, and the serial program corresponding to each branch in the probabilistic program;
- and
- execute the transformed instructions to generate samples,
- wherein the generated samples (i) are executed with a randomly selected control flow or a control flow selected from among previously unselected control flows for statistical model analysis, and (ii) logically contribute to calculation of an output of the serial program, thereby improving sampling efficiency of the probabilistic program.
A sampling method executed by a computer including a memory and a processor, the method comprising:
- transforming instructions of a serial program of a probabilistic program by using an internal state in reverse order of the serial program, the internal state being determined based on an instruction to be transformed, and the serial program corresponding to each branch in the probabilistic program;
- and
- executing the transformed instructions to generate samples,
- wherein the generated samples (i) are executed with a randomly selected control flow or a control flow selected from among previously unselected control flows for statistical model analysis, and (ii) logically contribute to calculation of an output of the serial program, thereby improving sampling efficiency of the probabilistic program.
Dependencies are inferred from the claim text.