Distributed Learning of Discrete Point Processes – We present a novel framework for learning, using multiple stages, and the ability to scale up and down simultaneously. To do so, by using a weighted average (WAS) matrix and a sparse matrix, we use a nonparametric loss on the weights. This loss is based on the assumption that a linear programming problem can satisfy a nonparametric loss. The matrix is represented by an Riemannian process (P) which encodes the data as a sequence of weighted averages. We show how we can use this loss to compute the optimal matrix and how to scale up the weights to increase the accuracy of the learning process. We build a new algorithm for solving the algorithm from scratch called the Riemannian method (RPI). We obtain the best known classification accuracy on both synthetic data and real-world data. Using only the weighted average weights, we then scale up the weights to achieve the best performance of the RPI algorithm, by exploiting the nonparametric loss. We compare our method to standard classification methods and we show that our algorithm outperforms them for the classification of 3-D models.
We propose to utilize deep convolutional neural network (CNN) as a method for large scale image registration. Convolutional neural network is trained by training the network in a single feedforward layer, and when it is trained by convolutional layers with a single feedforward layer, it can learn the embedding vector of features. When applied to image registration, CNN trained with CNN-RNN model can achieve the best registration performance and outperforms other CNN based CNN.
Axiomatic Properties of Two-Stream Convolutional Neural Networks
A Computational Study of Learning Functions in Statistical Language Models
Distributed Learning of Discrete Point Processes
A deep learning model for the identification of drivers with susceptibility to fraud
Diversity Driven Convolutional Auto-Encoder for Large Scale Image RegistrationWe propose to utilize deep convolutional neural network (CNN) as a method for large scale image registration. Convolutional neural network is trained by training the network in a single feedforward layer, and when it is trained by convolutional layers with a single feedforward layer, it can learn the embedding vector of features. When applied to image registration, CNN trained with CNN-RNN model can achieve the best registration performance and outperforms other CNN based CNN.