MF24-T03 AI: How activation functions can change your life… or at least improve your model (engl.)
Have you ever wondered what makes a deep-learning model work? Do you ever feel overwhelmed by the endless possibilities to improve your model and don’t know in which direction to take it, especially a model with as many hyperparameters and functions to choose from as a neural network? That was – and occasionally still is – me and many fellow artificial intelligence professionals. While it is true that engineering a neural network is often a heuristic process, and you have to test many models before finding a satisfiable one, a better understanding of the different components of a network and their mathematical properties, specifically activation functions, and taking a deep dive into what happens to them in backpropagation, can improve your sense of understanding of what your model is doing, and help you make informed changes in your quest for a model that fits your requirements. In this course, I will share with you the basic properties of different activation functions and their behaviour within a neural network, but also, and most importantly, show you how you can find out these properties by yourself and learn to play with a neural network with a better understanding of your moves.
Credit Points (ECTS): –
Dipl.-Ing. Maimouna Kebbou
I am a Computer Science Engineer and a Master student in Artificial Intelligence Engineering in my last semester. I have been working on AI projects for 5+ years both in academia and in various companies. I love to see how AI can improve processes in any field and how it can create new opportunities that were unimaginable a few years ago, and I am honored to feel part of it as an AI engineer.
Melde dich an, wenn du bereits ein Konto bei uns hast.