Consider a staff of medical professionals using a neural network to detect cancer in mammogram pictures. Even if this device-learning design looks to be carrying out effectively, it may be focusing on impression capabilities that are accidentally correlated with tumors, like a watermark or timestamp, relatively than actual indications of tumors.
To exam these models, researchers use “feature-attribution techniques,” strategies that are meant to notify them which components of the image are the most essential for the neural network’s prediction. But what if the attribution approach misses capabilities that are critical to the product? Given that the researchers never know which options are critical to get started with, they have no way of knowing that their analysis approach is not efficient.
To assist clear up this problem, MIT scientists have devised a method to modify the initial facts so they