Researchers to unveil New Programs written in Probabilistic Programming Language
In June, researchers will be unveiling programs written in probabilistic programming language at the Computer Vision and Pattern Recognition conference.
Under this pattern, task can be complete using just 50 lines of code. However, the same task is completed thousands lines of code by conventional systems.
Experts said due to advancement in machine-learning applications, the development of the new programming languages has thrived. MIT researchers will present programs written in their new code Picture. They are the ones that have been applied to standard computer-vision tasks.
Two plus points were noticed about these programs. They took much less code to write. Secondly, in some tasks, they actually performed better than earlier systems. The error rate was found to be between 50 and 80% lower than prior systems.
For the first time, the researchers have been introducing probabilistic programming in the vision area. "The whole hope is to write very flexible models, both generative and discriminative models, as short probabilistic code, and then not do anything else", stated Tejas Kulkarni, an MIT graduate student in brain and cognitive sciences.
The researchers have used inverse graphics. Kulkarni was of the view that probabilistic programming may ease re-writing code across different problems.