Computer scientists find the limits of the research algorithm
[ad_1]
Many aspects of modern applied research is based on a crucial algorithm called gradient drop. It is a process called optimizing the function of finding the largest or smallest values of a given mathematical function. It can be used to calculate the best way to manufacture a product and calculate the best way to give shifts to employees.
Despite its widespread use, researchers have never fully understood the situation in which the algorithm struggles the most. Now, the new work explains, setting this drop in gradients, at the heart, addresses a basic computational problem. The new result places limits on the type of performance that researchers can expect from the technique in certain applications.
“It’s the hardest case worth knowing,” he said Paul Goldberg Oxford University, author of the work John Fearnley and Rahul Savani University of Liverpool and Alexandros Hollender Oxford. The result received a Best Paper Award in June of the year Symposium on computer theory.
You can imagine a function as a landscape, where the elevation of the earth is equal to the value of the function (“gain”) in that particular place. The gradient drop seeks the minimum local of the function by looking for the direction of the hardest rise in a given location and looking down from there. The slope of the landscape is called the gradient, hence the gradient drop.
Gradient drop is a key tool in modern applied research, but there are many common problems that do not work well. But prior to this research, there was no definite understanding of what causes the fallout struggle and when — other areas of computing known as computational complexity theory helped respond.
“A lot of the work on the gradient drop wasn’t talking to the complexity theory,” he said Costis Daskalakis Massachusetts Institute of Technology.
Computational complexity is the study of the resources needed to solve or verify solutions to different computational problems, often computational time. Researchers classify problems into different classes, with all problems in the same class sharing some basic computational features.
To give an example — which is important for the new role — imagine that there are more people than houses and that everyone lives in the house. They give you a phone book with the names and addresses of all the people in the village, and ask them to find two people who live in the same house. You know you can find the answer, because there are more people than houses, but it needs a bit of a look (especially if they don’t share a last name).
This question refers to a complexity class called TFNP, abbreviated as “a function without a deterministic polynomial function”. It is a collection of all the computing problems that have guaranteed solutions and can quickly verify the correctness of the solutions. The researchers focused on the intersection of the two subsets within the TFNP.
The first subset is called PLS (polynomial local search). This is a collection of problems that involve finding the minimum or maximum value of a function in a particular region. It is ensured that these problems have answers that can be found through fairly straightforward reasoning.
One problem that falls into the PLS category is the task of planning a route, which allows you to visit a fixed number of cities with the shortest possible travel distance, as you can change the trip by changing the order of pairs in any city respectively. biran. It’s easy to calculate the length of any proposed route and, with the limitation of ways to adapt the route, it’s easy to see what changes shorten the trip. You make sure you find a route that you can’t improve with an acceptable move, the local minimum.
The second subset of the problem is PPAD (arguments for polynomial parity in directed graphs). These problems have solutions that arise from a more complicated process called Brouwer’s fixed point theorem. The theorem states that for any function to continue, it is certain that there will be a point that leaves the function unchanged – a fixed point, as is well known. It is true in everyday life. If a glass is mixed with water, the theorem guarantees that there must be a particle of water that will end up in the same place where it started.
[ad_2]
Source link