Sparse Sparse Coding for Deep Neural Networks via Sparsity Distributions – We propose a novel deep sparse coding method that is based on learning with a linear sparsity of the neural network features. Specifically, we propose a supervised supervised learning algorithm that learns a sparse coding model of the network features, which we call an adaptive sparse coding process (ASCP). Our method uses a linear regularization term to learn a sparse coding model of the network features. While our method learns a sparse coding model from the sparsity of network features, we also propose a linear sparsity term that is directly derived from spatial data sources. In this paper, we illustrate the proposed method through a simulated, real-world task, and show that our sparse coding algorithm outperforms state-of-the-art sparse coding methods in terms of accuracy.

Recently, the success of deep learning and deep learning-based generative adversarial neural networks (GANs) have led to a plethora of potential applications of the machine learning method for this task. In this paper, we investigate the possibility of deploying CNNs into a variety of applications; from learning to recognize faces using images of faces with different faces to pose-based models leveraging motion capture for face recognition. First we build a CNN trained to recognize faces from a large amount of data from a public database. After that, the CNN trained to solve a pose-based pose retrieval system on images of faces which are difficult to obtain at face recognition time. We demonstrate that our method successfully retrieved 3,732 face images of 8,929 subjects from the public public database. After extracting 3,767 face images from the database, our method is able to obtain a high recognition rate and successfully achieve good recognition rates. Our method significantly outperforms recent state-of-the-art pose and pose retrieval method.

Formalizing the Semi-Boolean Rule in Probability Representation

An Online Clustering Approach to Optimal Regression

Sparse Sparse Coding for Deep Neural Networks via Sparsity Distributions

A hybrid algorithm for learning sparse and linear discriminant sequences

Action Recognition with 3D CNN: Onsets and TransformationsRecently, the success of deep learning and deep learning-based generative adversarial neural networks (GANs) have led to a plethora of potential applications of the machine learning method for this task. In this paper, we investigate the possibility of deploying CNNs into a variety of applications; from learning to recognize faces using images of faces with different faces to pose-based models leveraging motion capture for face recognition. First we build a CNN trained to recognize faces from a large amount of data from a public database. After that, the CNN trained to solve a pose-based pose retrieval system on images of faces which are difficult to obtain at face recognition time. We demonstrate that our method successfully retrieved 3,732 face images of 8,929 subjects from the public public database. After extracting 3,767 face images from the database, our method is able to obtain a high recognition rate and successfully achieve good recognition rates. Our method significantly outperforms recent state-of-the-art pose and pose retrieval method.