Training Convolutional Neural Networks with the Forward-Forward algorithm

📅 2023-12-22
🏛️ arXiv.org
📈 Citations: 11
Influential: 2
📄 PDF
🤖 AI Summary
This work addresses the challenge of adapting the Forward-Forward (FF) algorithm—originally designed for fully connected networks—to convolutional neural networks (CNNs). We propose the first purely forward FF training framework for CNNs, eliminating backpropagation entirely. Our core innovation is a spatially expanded label mechanism that aligns class labels with local receptive fields, enabling pixel-wise forward discriminative learning; we further integrate Class Activation Maps (CAM) for interpretable feature visualization. On MNIST, our method achieves 99.16% accuracy—on par with standard backpropagation baselines. Crucially, we present the first successful FF-CNN evaluations on CIFAR-10 and CIFAR-100, demonstrating its capacity to autonomously learn semantically coherent spatial features. Ablation studies validate the efficacy of each component. This work establishes a novel paradigm for training vision models without gradient-based optimization, advancing the frontier of biologically plausible and computationally efficient deep learning.
📝 Abstract
The recent successes in analyzing images with deep neural networks are almost exclusively achieved with Convolutional Neural Networks (CNNs). The training of these CNNs, and in fact of all deep neural network architectures, uses the backpropagation algorithm where the output of the network is compared with the desired result and the difference is then used to tune the weights of the network towards the desired outcome. In a 2022 preprint, Geoffrey Hinton suggested an alternative way of training which passes the desired results together with the images at the input of the network. This so called Forward Forward (FF) algorithm has up to now only been used in fully connected networks. In this paper, we show how the FF paradigm can be extended to CNNs. Our FF-trained CNN, featuring a novel spatially-extended labeling technique, achieves a classification accuracy of 99.16% on the MNIST hand-written digits dataset. We show how different hyperparameters affect the performance of the proposed algorithm and compare the results with CNN trained with the standard backpropagation approach. Furthermore, we use Class Activation Maps to investigate which type of features are learnt by the FF algorithm.
Problem

Research questions and friction points this paper is trying to address.

Extending Forward-Forward algorithm to Convolutional Neural Networks for image analysis
Developing spatial labeling strategies to prevent shortcut solutions in training
Enabling biologically inspired learning for neuromorphic computing systems
Innovation

Methods, ideas, or system contributions that make the work stand out.

Forward-Forward algorithm replaces backpropagation in CNNs
Spatial label strategies use Fourier and morphology patterns
Local goodness function enables biologically plausible learning
🔎 Similar Papers
No similar papers found.
R
Riccardo Scodellaro
Translational Molecular Imaging, Max Planck Institute for Multidisciplinary Sciences, Hermann-Rein Straße 3, 37075 Göttingen, Germany
A
A. Kulkarni
Translational Molecular Imaging, Max Planck Institute for Multidisciplinary Sciences, Hermann-Rein Straße 3, 37075 Göttingen, Germany
F
Frauke Alves
Translational Molecular Imaging, Max Planck Institute for Multidisciplinary Sciences, Hermann-Rein Straße 3, 37075 Göttingen, Germany; Department of Haematology and Medical Oncology, University Medical Center Göttingen, Robert Koch-Straße 40, 37075 Göttingen, Germany; Institute for Diagnostic and Interventional Radiology, University Medical Center Göttingen, Robert Koch-Straße 40, 37075 Göttingen, Germany
M
Matthias Schröter
Institute for Diagnostic and Interventional Radiology, University Medical Center Göttingen, Robert Koch-Straße 40, 37075 Göttingen, Germany; Max Planck Institute for Dynamics and Self-Organization, Am Faßberg 17, 37075 Göttingen, Germany