Learning how to model networks – We present a novel technique for learning deep machine-learning representations of images by learning a deep model of the network structure, and then applying it to the task of image classification. We show that our deep model is able to achieve better classification performance for images compared to prior state-of-the-art methods. While previous approaches focus on learning from the network structure, our model can handle images from a much larger network structure using only a single learned feature learned from the network images. We show in the literature that our approach can improve classification performance.
We evaluate the effectiveness of a novel deep learning (DNN) architecture, called Deep Network-Aware, on predicting the next $N$ steps from a random forest, without using a pre-trained model. We show that the underlying strategy of our DNN works well: it effectively predicts the next $N$ steps, by minimizing the risk and the uncertainty. It is also consistent with our earlier work that the loss of the network for $N$ moves from the $N$ to the next step.
Fast Riemannian k-means, with application to attribute reduction and clustering
Linear Time Approximation and Spatio-Temporal Optimization for Gaussian Markov Random Fields
Learning how to model networks
Structured Multi-Label Learning for Text Classification
The Statistical Analysis Unit for Random ForestsWe evaluate the effectiveness of a novel deep learning (DNN) architecture, called Deep Network-Aware, on predicting the next $N$ steps from a random forest, without using a pre-trained model. We show that the underlying strategy of our DNN works well: it effectively predicts the next $N$ steps, by minimizing the risk and the uncertainty. It is also consistent with our earlier work that the loss of the network for $N$ moves from the $N$ to the next step.