Inserting machine learning in physics courses: the case of rolling on an inclined plane

In an increasingly data-oriented science, the use of automatic computational methods is progressively indispensable. In this context, it becomes important to expose undergraduate physics students to artificial intelligence and machine learning methodologies. In this work we propose a way to use such methods in physics, solving the didactic problem of rolling on an inclined plane. We introduce the main concepts of machine learning techiniques and measure the travel time of different objects (rim, disk and sphere) for an initial height and tilt angle. Based on these data, we used classification models capable of predicting the object that was dropped with an accuracy of 83%, and regression models which were able to predict the average speed of the object that was rolled with mean absolute error of 1.4 cm s⁻¹. We also show that this didactic model is instructive because it allows a direct comparison with physical models and serves as a discussion of the meaning of teaching physics to the computer.

Keywords:
Artificial Intelligence; Machine Learning; Inclined Plane