Check out RSS, or use RSS reader to subscribe this item
Confirmation
Authentication email has already been sent, please check your email box: and activate it as soon as possible.
You can login to My Profile and manage your email alerts.
Sponsored by the Center for Science and Technology Development of the Ministry of Education
Supervised by Ministry of Education of the People's Republic of China
Convolutional neural networks (ConvNets) is multi-stages trainable architecture that can learn invariant features in recognition. Applications of ConvNets in 'Big Data' are always limited to some challenges: 1) the labeled data is scarce and the labeled data is abundant; 2) tedious training procedure is required frequently with updated training samples. In this work, an efficient principle component analysis(PCA) based pre-training strategy has been introduced to reduce the high computational cost of kernel training in ConvNets, and to make the system be more robust to insufficient labeled training data. Two datasets MNIST and VLOGO are employed to validate the proposed work. The classification experiments results have demonstrated that the proposed pre-training ConvNets is able to accelerate the training procedure and reduce the requirement of sufficient labeled training samples.
Keywords:signal and information processing; convolutional neural networks; pre-training; recognition; MNIST; vehicle logo