Learning similarity functions between image pairs with deep neural networks yields highly correlated activations of large embeddings. In this work, we show how to improve the robustness of embeddings by exploiting independence in ensembles. We divide the last embedding layer of a deep network into an embedding ensemble and formulate training this ensemble as an online gradient boosting problem....

Digital watermarking is referred to a method used for copyright protection and authentication. In this paper, we present a method of nested digital watermark embedding and extraction in which a nested watermark (a watermark inside another watermark) is embedded into the main image. This concept of nested watermarking is used to increase the watermark embedding capacity. In this method, a waterm...

This paper addresses a novel steganography method for images. Most statistical steganalysis algorithms are strong to defeat previous steganography algorithms. RS steganalysis and pixel difference histogram analysis are two well-known statistical steganalysis algorithms which detect non-random changes caused by embedding a secret message into cover image. In this paper, we first explain how two ...

Knowledge embedding, which projects triples in a given knowledge base to d-dimensional vectors, has attracted considerable research efforts recently. Most existing approaches treat the given knowledge base as a set of triplets, each of whose representation is then learned separately. However, as a fact, triples are connected and depend on each other. In this paper, we propose a graph aware know...

Embedding algorithms are a method for revealing low dimensional structure in complex data. Most embedding algorithms are designed to handle objects of a single type for which pairwise distances are specified. Here we describe a method for embedding objects of different types (such as authors and terms) into a single common Euclidean space based on their co-occurrence statistics. The joint distr...

Total embedding distributions have been known for only a few classes of graphs. In this paper the total embedding distributions of the cacti and the necklaces are obtained. Furthermore we obtain the total embedding distributions of all graphs with maximum genus 1 by using the method of this paper.

Commute time embedding involves computing eigenfunctions of the graph Laplacian matrix. Spectral decomposition requires computational burden proportional to 3 ( ) O n , which may not be suitable for large scale dataset. This paper proposes computationally efficient commute time embedding by applying Nyström method to the normalized graph Laplacian. The performance of the proposed algorithms is ...