Introduction to Adversarial Machine Learning
Image credit: MCS_SSCAbstract
With all the advances in machine learning and especially in deep learning, you may think that these models are robust and almost perfect in at least the easier tasks such as identifying animals. Unfortunately you’re wrong. While these models have shown great achievements in many tasks, even better than humans, they’re very vulnerable to a family of attacks called Adversarial Attacks, an unsolved problem and an active field of research in machine learning, which we are going to talk about in this presentation
Adversarial Machine Learning
Arian Amani
Machine Learning Scientist
I am a Machine Learning Scientist at AI VIVO and a Data Scientist at the Wellcome Sanger Institute. My work is at the intersection of computational biology and drug discovery, where I develop deep generative and foundation models for molecules and cells. I specialize in molecule generation and single-cell perturbation modeling using advanced techniques like VAEs, Diffusions, Transformers, and Flow Matchings. I’m passionate about building AI methods that accelerate target discovery and therapeutic design.