Quanta Magazine March 8, 2021
For solving linear systems researchers at the Georgia Institute of Technology have developed a technique that employs an enhanced version of the iterated guessing strategy: Instead of making just a single guess, their algorithm makes many guesses in parallel. This approach speeds up the search. The key feature to the algorithm’s ultimate success is that it makes the three initial guesses at random. Randomness ensures you do not accidentally end up biasing your search toward one part of the problem, potentially neglecting the space where the actual solution lies. They showed that the different strands of random guesses also work together, including any guess that is in fact the answer to the problem. The coordinated randomness account for the exact values of the guesses themselves and cover all the potential guesses that lie between them. The new technique can find the solution within the matrix faster than it could in a matrix with the same number of entries…read more. Open Access TECHNICAL ARTICLEÂ