Neural Network Recognizer for Hand-Written Zip Code Digits

By John S. Denker, W. Gardner, Hans Peter Graf, D. Henderson, Richard E. Howard, Wayne E. Hubbard, L. Jackel, Henry S. Baird, Isabelle M Guyon. 1988

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

Learning Robust Representations by Projecting Superficial Statistics Out (2018, methodology): “In order to search the old computer vision techniques for a method that can extract more textural information and less semantic information, we experimented with three classifcal computer vision techniques: SURF (Bay et al., 2006), LBP (He & Wang, 1990), and GLCM (Haralick et al., 1973) on several d”
Unsupervised Pixel-Level Domain Adaptation with Generative Adversarial Networks (2016, methodology): “We evaluate our method on object classiﬁcation datasets used in previous work 1 , including MNIST, MNIST-M [14], and USPS [10] as well as a variation of the LineMod dataset [22, 46], a standard for object instance recognition and 3D pose estimation, for which we have synthetic and real data.”
Broad Learning System: An Effective and Efficient Incremental Learning System Without the Need for Deep Architecture (2018, methodology): “Conventional methods for training SLFN are gradient-descent-based learning algorithms [22], [23].”
Backpropagation Applied to Handwritten Zip Code Recognition (1989, methodology): “Unlike previous results reported by our group on this problem (Denker et al. 1989), the learning network is directly fed with images, rather than feature vectors, thus demonstrating the ability of backpropagation networks to deal with large amounts of low-level information.”
Progressive Feature Alignment for Unsupervised Domain Adaptation (2018, background): “In particular, the images in MNIST and SVHN are grey, and are of size 28 × 28 and 16 × 16, respectively; USPS consists of color images of size 32×32, and there are often more than one digit in one image.”
Handwritten zip code recognition with multilayer networks (1990, result): “This is in contrast with our earlier work [ 2 ] where the first few layers of connections were chosen by hand and were not subject to learning.”
Causal Generative Domain Adaptation Networks (2018, methodology): “Following CoGAN [33, 34], we use the standard training-test splits for both MNIST and USPS.”
A comprehensive evaluation of random vector functional link networks (2016, methodology): “Conventional methods for training SLFN are back-propagation based learning algorithms [10, 7].”
Learning Models with Uniform Performance via Distributionally Robust Optimization (2018, methodology): “We construct our training set as a mixture of MNIST hand-written digits [33] (majority population) and type-written digits consisting of different fonts [31] (minority population).”
Learning to Diversify for Single Domain Generalization (2021, background): “To further demonstrate the effectiveness of the proposed method, we use t-SNE [41] to visualize the distribution of the unseen target features in the digits dataset (i.e., SVHN, SYN, USPS and MNIST-M).”
Unsupervised Domain Adaptation with Similarity Learning (2017, background): “We show results in three domain adaptation datasets: Digits, which contains digit images from difference domains (MNIST [32], USPS [13] and MNIST-M [15]), Office-31 [45], which contains images of office objects in three different domains and VisDA [42], a large-scale dataset focused on simulation-to”
Learning to Learn Single Domain Generalization (2020, background): “Datasets and settings: (1) Digits consists of ﬁve sub-datasets: MNIST [19], MNIST-M [8], SVHN [34], SYN [8], and USPS [5], and each of them can be viewed as a different domain.”