Learning algorithms for classification: A comparison on handwritten digit recognition

By Yann LeCun, L. Jackel, L. Eon Bottou, Corinna Cortes, John S. Denker, Harris Drucker, Isabelle M Guyon, Urs Muller, Eduard Säckinger, Patrice Y. Simard and 1 more. 1995

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

Xception: Deep Learning with Depthwise Separable Convolutions (2016, methodology): “The history of convolutional neural network design started with LeNet-style models [10], which were simple stacks of convolutions for feature extraction and max-pooling operations for spatial sub-sampling.”
Multi-column deep neural networks for image classification (2012, background): “2% error rate of humans on this task [21].”
Scalable, High-Quality Object Detection (2014, background): “Deep neural network architectures with repeated convolution and pooling layers [7, 13] have more recently become the dominant approach for large-scale and high-quality recognition and detection.”
Neural Cleanse: Identifying and Mitigating Backdoor Attacks in Neural Networks (2019, background): “The goal is to recognize 10 hand-written digits (0-9) in gray-scale images [23].”
BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain (2017, methodology): “Our first set of experiments uses the MNIST digit recognition task [37], which involves classifying grayscale images of handwritten digits into ten classes, one corresponding to each digit in the set [0, 9].”
Transformation Invariance in Pattern Recognition-Tangent Distance and Tangent Propagation (1996, background): “The most computationally and memory-efficient algorithms generally involve learning (LeCun et al., 1995).”
Adversarial Diversity and Hard Positive Generation (2016, methodology): “We generated one million new examples with InfiMNIST and tested them on three differently trained LeNet [13] networks to identify adversarial examples among these images.”
Deep Networks from the Principle of Rate Reduction (2020, background): “The resulting network will be a deep convolutional network that shares the same multi-channel convolution structures as conventional CNNs for 2D images (LeCun et al., 1995; Krizhevsky et al., 2012).”
Learning Continuous Attractors in Recurrent Networks (1997, methodology): “Experiments were conducted with images of handwritten digits from the USPS database described in [12].”
Malware traffic classification using convolutional neural network for representation learning (2017, methodology): “The size of preprocessed dataset and the size of each image are both very similar to classic MNIST dataset [21], so a CNN architecture similar to LeNet-5 [22] was used but with much more channels in each convolution layer.”
Prospects for Theranostics in Neurosurgical Imaging: Empowering Confocal Laser Endomicroscopy Diagnostics via Deep Learning (2018, background): “(37) for handwritten digit recognition, was trained on the training dataset until it produced descent classification results on validation images.”
Shape representation and classification using the Poisson equation (2004, result): “Our results are superior to many of the algorithms tested in the benchmark [15], including RBF, Neural nets, PCA, and linear classifiers, but are inferior to the best results published for boosted LeNet-4 [15] ( 0.”
Convolutional Neural Network Committees for Handwritten Character Classification (2011, methodology): “Our committee-based classifiers of isolated handwritten characters are the first on par with human performance [27], [28], and can be used as basic building blocks of any OCR system (all our results were achieved by software running on powerful yet cheap gaming cards).”
HAST-IDS: Learning Hierarchical Spatial-Temporal Features Using Deep Neural Networks to Improve Intrusion Detection (2018, background): “An interesting finding is that applying a similar deep neural network architecture to classify the MNIST hand-written digit dataset also achieves very high accuracy [32].”
Assessing Threat of Adversarial Examples on Deep Neural Networks (2016, methodology): “The described experiments were run both on the raw LeNet [15] network, and on a fine-tuned network.”