{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/biologically-plausible-learning-algorithms","title":"Biologically-plausible learning algorithms can scale to large datasets","arxiv_id":"1811.03567","date":"2018-11-08","proceeding":"ICLR 2019 5","authors":["Will Xiao","Honglin Chen","Qianli Liao","Tomaso Poggio"],"abstract":"The backpropagation (BP) algorithm is often thought to be biologically\nimplausible in the brain. One of the main reasons is that BP requires symmetric\nweight matrices in the feedforward and feedback pathways. To address this\n\"weight transport problem\" (Grossberg, 1987), two more biologically plausible\nalgorithms, proposed by Liao et al. (2016) and Lillicrap et al. (2016), relax\nBP's weight symmetry requirements and demonstrate comparable learning\ncapabilities to that of BP on small datasets. However, a recent study by\nBartunov et al. (2018) evaluate variants of target-propagation (TP) and\nfeedback alignment (FA) on MINIST, CIFAR, and ImageNet datasets, and find that\nalthough many of the proposed algorithms perform well on MNIST and CIFAR, they\nperform significantly worse than BP on ImageNet. Here, we additionally evaluate\nthe sign-symmetry algorithm (Liao et al., 2016), which differs from both BP and\nFA in that the feedback and feedforward weights share signs but not magnitudes.\nWe examine the performance of sign-symmetry and feedback alignment on ImageNet\nand MS COCO datasets using different network architectures (ResNet-18 and\nAlexNet for ImageNet, RetinaNet for MS COCO). Surprisingly, networks trained\nwith sign-symmetry can attain classification performance approaching that of\nBP-trained networks. These results complement the study by Bartunov et al.\n(2018), and establish a new benchmark for future biologically plausible\nlearning algorithms on more difficult datasets and more complex architectures.","url_abs":"http://arxiv.org/abs/1811.03567v3","url_pdf":"http://arxiv.org/pdf/1811.03567v3.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"biologically-plausible-learning-algorithms","repo_url":"https://github.com/willwx/sign-symmetry","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"biologically-plausible-learning-algorithms","repo_url":"https://github.com/jsalbert/biotorch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[{"task_slug":null,"task_name":"Biologically-plausible Training"}],"methods":[{"method_slug":"1x1-convolution","method_name":"1x1 Convolution"},{"method_slug":"average-pooling","method_name":"Average Pooling"},{"method_slug":"batch-normalization","method_name":"Batch Normalization"},{"method_slug":"bottleneck-residual-block","method_name":"Bottleneck Residual Block"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"fpn","method_name":"FPN"},{"method_slug":"focal-loss","method_name":"Focal Loss"},{"method_slug":"global-average-pooling","method_name":"Global Average Pooling"},{"method_slug":"grouped-convolution","method_name":"Grouped Convolution"},{"method_slug":"kaiming-initialization","method_name":"Kaiming Initialization"},{"method_slug":"local-response-normalization","method_name":"Local Response Normalization"},{"method_slug":"max-pooling","method_name":"Max Pooling"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"residual-block","method_name":"Residual Block"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"retinanet","method_name":"RetinaNet"},{"method_slug":"sgd-with-momentum","method_name":"SGD with Momentum"},{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1811.03567","atlas_url":"https://app.syntology.ai/?focus=1811.03567","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}