1.WHY THEY CALLED ML AS “BLACK BOX”?
These powerful machine learning models are typically based on artificial neural networks, which can have millions of nodes processing data to make predictions. Because of their complexity, researchers often refer to these models as “black boxes.” Even the scientists building the models don’t understand everything that’s going on under the hood.
2.WHAT ARE THE OTHER EXAMPLES OF ML?
Other examples of machine learning on graphs are traffic routing, chip design, and recommender systems. Designing these models is made even more difficult by the fact that the data used to train these models is often different from the data the models see in the field. Presumably the model was trained using a small molecular graph or traffic network.
3.WHICH KIND OF EXPECTATIONS RESEARCHERS HAVE ON ML?
These models reuse the learned representations later when refining them for specific tasks. Ideally, researchers want the model to learn as much as possible during pre-training so that it can apply that knowledge to downstream tasks. In machine learning, we often see something happening and try to understand it theoretically. That’s a big challenge. We want to build an understanding that matches what we actually see so we can do better. We are just beginning to understand this.