Dresden 2026 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz
AKPIK 2: Machine Learning Prediction and Optimization Tasks
AKPIK 2.3: Talk
Tuesday, March 10, 2026, 10:00–10:15, BEY/0127
Training convolutional neural networks with the forward - forward algorithm — •Matthias Schröter1,2, Frauke Alves3, and Riccardo Scodellaro3 — 1Institute for Diagnostic and Interventional Radiology, University Medical Center Göttingen, Robert Koch-Straße 40, 37075 Göttingen, Germany — 2Max Planck Institute for Dynamics and Self-Organization, 37075 Göttingen, Germany — 3Translational Molecular Imaging, Max Planck Institute for Multidisciplinary Sciences, 37075 Göttingen, Germany.
Recent successes in image analysis with deep neural networks are achieved almost exclusively with Convolutional Neural Networks (CNNs) trained using the backpropagation (BP) algorithm. In a 2022 preprint, Geoffrey Hinton proposed the Forward - Forward (FF) algorithm as a biologically inspired alternative, where positive and negative examples are jointly presented to the network and training is guided by a locally defined goodness function. Here, we extend the FF paradigm to CNNs. This talk compares FF and BP training across different datasets (MNIST, CIFAR 10, CIFAR 100) discusses different optimization strategies, and provides insights into the inner workings of FF trained networks using Class Activation Maps.
Keywords: Convolutional Neural Networks; Forward-Forward training; backpropagation; neuromorphic computing