Effective Training of a Neural Network Character Classifier for Word Recognition

By Larry S. Yaeger, R. Lyon, B. J. Webb. 1996

How researchers see this paper

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

Citations in the corpus

APAC: Augmented PAttern Classification with Neural Networks (2015, background): “Handwritten character/digit recognition has been an important problem for both industrial applications and algo-rithm benchmarking for a quarter century [23, 10, 17, 11, 2, 3].”
A Bayesian Data Augmentation Approach for Learning Deep Models (2017, methodology): “[30] proposed the “stroke warping" technique for word recognition, which adds small changes in skew, rotation, and scaling into the original word images.”
Transformation Pursuit for Image Classification (2014, unknown): “In this work, we focus on such transformations, as they have been shown to increase classification accuracy, especially on simple tasks such as digit recognition [15, 9, 10, 34], and, more recently, on ImageNet [14].”
Improving Deep Learning using Generic Data Augmentation (2017, methodology): “A potential solution is to make use of Data Augmentation (DA) [Yaeger et al., 1996], which is a regularization scheme that artificially inflates the data-set by using label preserving transformations to add more invariant examples.”
Improving Deep Learning with Generic Data Augmentation (2018, methodology): “A potential solution is to use Data Augmentation (DA) [5], which is a regularization scheme that artiﬁcially inﬂates the data-set by using label preserving transformations to add more invariant examples.”
Invariance-inducing regularization using worst-case transformations suffices to boost accuracy and spatial robustness (2019, methodology): “On the other hand, augmentation-based methods [3, 46] constitute an alternative approach to encourage desired invariances on transformation sets.”
Best practices for convolutional neural networks applied to visual document analysis (2003, background): “If the data is scarce and if the distribution to be learned has transformation-invariance properties, generating additional data using transformations may even improve performance [6].”
Pretraining Convolutional Neural Networks for Visual Recognition (2016, background): “If a training set is scarce and has some transformation-invariance properties, generating additional data may be useful [39, 40].”
Understanding and Mitigating the Tradeoff Between Robustness and Accuracy (2020, methodology): “This is similar to random data augmentation in images (Yaeger et al., 1996; Krizhevsky et al., 2012), where inputs are perturbed in a way that preserves the label.”
Regularization for Deep Learning: A Taxonomy (2017, background): “Rigid and elastic image transformation (Baird, 1990; Yaegger et al., 1996; Simard et al., 2003; Ciresan et al., 2010) p(θ) Input Random Training”