Efficient Deep Feature Learning and Extraction via StochasticNets
This work addresses the need for more efficient deep learning models for researchers and practitioners in computer vision, offering incremental improvements in speed and accuracy with sparse connectivity.
The paper tackles the problem of inefficient neural connectivity in deep neural networks for feature learning and extraction by proposing StochasticNets, which use stochastic connectivity to form sparsely-connected networks. The result is that features learned with StochasticNets, using fewer connections than conventional networks, achieve better or comparable classification accuracy, with a relative test error decrease of ~4.5% on STL-10 and ~1% on SVHN, and allow for faster feature extraction in embedded applications.
Deep neural networks are a powerful tool for feature learning and extraction given their ability to model high-level abstractions in highly complex data. One area worth exploring in feature learning and extraction using deep neural networks is efficient neural connectivity formation for faster feature learning and extraction. Motivated by findings of stochastic synaptic connectivity formation in the brain as well as the brain's uncanny ability to efficiently represent information, we propose the efficient learning and extraction of features via StochasticNets, where sparsely-connected deep neural networks can be formed via stochastic connectivity between neurons. To evaluate the feasibility of such a deep neural network architecture for feature learning and extraction, we train deep convolutional StochasticNets to learn abstract features using the CIFAR-10 dataset, and extract the learned features from images to perform classification on the SVHN and STL-10 datasets. Experimental results show that features learned using deep convolutional StochasticNets, with fewer neural connections than conventional deep convolutional neural networks, can allow for better or comparable classification accuracy than conventional deep neural networks: relative test error decrease of ~4.5% for classification on the STL-10 dataset and ~1% for classification on the SVHN dataset. Furthermore, it was shown that the deep features extracted using deep convolutional StochasticNets can provide comparable classification accuracy even when only 10% of the training data is used for feature learning. Finally, it was also shown that significant gains in feature extraction speed can be achieved in embedded applications using StochasticNets. As such, StochasticNets allow for faster feature learning and extraction performance while facilitate for better or comparable accuracy performances.