A quick method has appeared that allows AI to forget information. Now for this, researchers do not need to completely rewrite the data.
Now, in order to make the trained AI model forget some data, it is often necessary to retrain it - delete all the data, and then write it again, removing unnecessary elements. This calculation process can take up to a week.
Scientists at Stanford University have proposed ways to more efficiently remove records from AI models. Researchers are confident that their approach will save megawatts of energy and allow models to run faster.
According to hightech.fm , Stanford researchers have applied the principle of a machine learning algorithm called the k-means method. This allows scientists to automatically break down many elements into a pre-known number of clusters. The main idea is that at each iteration, the center of mass is recalculated for each cluster obtained in the previous step, then the vectors are again divided into clusters according to which of the new centers is closer in the selected metric.
So scientists grouped data sets and tested their editing on six data sets. So they classified handwritten numbers, hand gestures, and other information. They were able to remove a thousand data points from each set without data loss.
© All rights reserved. Citing to www.ict.az is necessary upon using news