Visual object classification by sparse convolutional neural networks

Abstract. A convolutional network architecture termed sparse convolutional neural network (SCNN) is proposed and tested on a real-world classification task (car classification). In addition to the error function based on the mean squared error (MSE), approximate decorrelation between hidden layer neurons is enforced by a weight orthogonalization mechanism. The aim is to obtain a sparse coding of the objects' visual appearance, thus removing the need for a dedicated feature selection stage. Working on unprocessed image data only, it is demonstrated that classification accuracies can be improved by the proposed method compared to purely MSE-trained SCNNs and fully-connected multilayer perceptron architectures.

• Type: Conference
• Year: 2006
• Where: ESANN
• Authors: Alexander Gepperth
• Keywords: Convolutional Network Architecture, Convolutional Neural Network, ESANN 2006, ESANN 2007, Real-world Classification Task