Large-scale Learning with SVM and Convolutional for Generic Object Categorization

By Fu Jie Huang, Yann LeCun. 2006

386 citations recorded by the source corpus.

Subfields: Artificial Intelligence, 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 Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Deeply-Supervised Nets (2014, methodology): “However, even though hierarchical and recursive networks [7, 10, 12] have shown great promise in automatically learning thousands or even millions of features for pattern recognition, there nonetheless remain many fundamental open questions about deep learning.”
Recent advances in convolutional neural networks (2015, background): “Low-rank matrix factorization has been exploited in a variety of contexts to improve the optimization problems.”
Deep Learning using Linear Support Vector Machines (2013, methodology): “The corresponding hidden variables of data samples are then treated as input and fed into linear (or kernel) SVMs (Huang & LeCun, 2006; Lee et al., 2009; Quoc et al., 2010; Coates et al., 2011).”
Learning Deep Energy Models (2011, methodology): “We highlight that our model is a fully connected model which does not assume any structure in the input data, unlike convolution based approaches which hold the current state-of-the-art performances on classification (97.2%, Huang & LeCun (2006); Le et al. (2010); Coates et al. (2011)).”
A Deep Structured Model with Radius–Margin Bound for 3D Human Activity Recognition (2015, methodology): “In the deep learning research, the combination of SVM and CNNs has been exploited [15] and obtained excellent results in object detection [12].”
Towards Scalable Representations of Object Categories: Learning a Hierarchy of Parts (2007, methodology): “Approaches that do build the layers by learning and are able to make a sufficient number of them (by starting with simple features) mostly design the parts by histogramming the local neighborhoods of parts of the previous layers [1] or by learning the neural weights based on the responses on previou”
High-dimensional and large-scale anomaly detection using a linear one-class SVM with deep learning (2016, methodology): “, 10(6) 10(4)) has been limited with SVMs [14].”
Large-Scale FPGA-based Convolutional Networks (2011, methodology): “In LeCun et al. (1998b) and Huang and LeCun (2006), pure supervised learning is used to update the parameters.”
Unsupervised Learning of Invariant Feature Hierarchies with Applications to Object Recognition (2008, methodology): “When multiple such levels are stacked, the resulting architecture is esse ntially identical to the Neocognitron [ 7], the Convolutional Network [13, 10], and the HMAX, or so-called “Standard Model” architecture [ 20, 17].”
Unsupervised Learning of Invariant Feature Hierarchies with Applications to Object Recognition (2007, methodology): “When multiple such levels are stacked, the resulting architecture is essentially identical to the Neocognitron [7], the Convolutional Network [13, 10], and the HMAX, or so-called “Standard Model” architecture [20, 17].”
Evaluation of Pooling Operations in Convolutional Architectures for Object Recognition (2010, methodology): “In order to keep our results comparable, we deliberately designed the networks to resemble the one by Huang and LeCun [7] for NORB and the one by Ahmed et al.”
Deep Convolutional Neural Networks for Image Classification: A Comprehensive Review (2017, methodology): “…Chopra, Hadsell, & LeCun, 2005); off-road obstacle avoidance (Muller, Ben, Cosatto, Flepp, & Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/neco_a_00990 by guest on 04 June 2021 LeCun, 2005); and generic object classification (LeCun, Huang, & Bottou, 2004; Huang & LeCun, 2006).”
Market impact analysis via deep learned architectures (2018, background): “Huang and LeCun [17] present a hybrid system where invariant image features are learned by a trained convolutional network, and object classification task is then conducted by a Gaussian kernel SVM based on learned features.”
Convolutional Neural Network Committees for Handwritten Character Classification (2011, background): “Recent CNN work focused on computer vision problems such as recognition of 3D objects, natural images and traffic signs [3]–[5], image denoising [6] and image segmentation [7].”
Selecting Receptive Fields in Deep Networks (2011, neutral): “, [4, 6, 9, 19]).”
A Tutorial on Energy-Based Learning (2006, background): “However, several authors have recently argued that convex loss functions are no guarantee for good performance, and that non-convex losses may in fact be easier to optimize than convex ones in practice, even in the absence of theoretical guarantees [36, 22].”