Nagoya Institute of Technology

About: Nagoya Institute of Technology is a education organization based out in Nagoya, Japan. It is known for research contribution in the topics: Thin film & Catalysis. The organization has 10766 authors who have published 19140 publications receiving 255696 citations. The organization is also known as: Nagoya Kōgyō Daigaku & Nitech.

Product of Experts for Statistical Parametric Speech Synthesis

Abstract: Multiple acoustic models are often combined in statistical parametric speech synthesis. Both linear and non-linear functions of an observation sequence are used as features to be modeled. This paper shows that this combination of multiple acoustic models can be expressed as a product of experts (PoE); the likelihoods from the models are scaled, multiplied together, and then normalized. Normally these models are individually trained and only combined at the synthesis stage. This paper discusses a more consistent PoE framework where the models are jointly trained. A training algorithm for PoEs based on linear feature functions and Gaussian experts is derived by generalizing the training algorithm for trajectory HMMs. However for non-linear feature functions or non-Gaussian experts this is not possible, so a scheme based on contrastive divergence learning is described. Experimental results show that the PoE framework provides both a mathematically elegant way to train multiple acoustic models jointly and significant improvements in the quality of the synthesized speech.

Geometry of deformed exponential families: Invariant, dually-flat and conformal geometries

Abstract: An information-geometrical foundation is established for the deformed exponential families of probability distributions Two different types of geometrical structures, an invariant geometry and a flat geometry, are given to a manifold of a deformed exponential family The two different geometries provide respective quantities such as deformed free energies, entropies and divergences The class belonging to both the invariant and flat geometries at the same time consists of exponential and mixture families Theq-families are characterized from the viewpoint of the invariant and flat geometries The q-exponential family is a unique class that has the invariant and flat geometries in the extended class of positive measures Furthermore, it is the only class of which the Riemannian metric is conformally connected with the invariant Fisher metric

One-step hydrothermal synthesis of V1-xWxO2(M/R) nanorods with superior doping efficiency and thermochromic properties

Abstract: Pure phase V1−xWxO2(M/R) nanorods with superb metal–insulator transition properties were obtained in tungsten (W)-doped level ranges from 0.5 to 2.0 at% using a one-step hydrothermal treatment without additional annealing steps. The assured level of W doping greatly promotes the grain growth of pure phase vanadium oxide (VO2)(M/R) and simultaneously the phase transition temperature (Tc) is depressed as much as 103 °C per at% W for the V1−xWxO2 nanorods when x = 1.0–2.0 at%. After mixing the pure phase VO2(M) (W-doped 0.5 at%, phase transition at 47 °C) nanorods with acrylic resin, the integrated visible transmission of the VO2 composite coating on glass is up to 60.6% and the integrated solar modulation efficiency is up to 10.3%. These results mean that the superior thermochromic property will greatly favor the practical application of VO2-based smart windows.

Speaker-Independent HMM-based Speech Synthesis System: HTS-2007 System for the Blizzard Challenge 2007

Abstract: This paper describes an HMM-based speech synthesis system developed by the HTS working group for the Blizzard Challenge 2007. To further explore the potential of HMM-based speech synthesis, we incorporate new features in our conventional system which underpin a speaker-independent approach: speaker adaptation techniques; adaptive training for HSMMs; and full covariance modeling using the CSMAPLR transforms.

How to Prove Impossibility Under Global Fairness: On Space Complexity of Self-Stabilizing Leader Election on a Population Protocol Model

Abstract: A population protocol is one of distributed computing models for passively-mobile systems, where a number of agents change their states by pairwise interactions between two agents. In this paper, we investigate the solvability of the self-stabilizing leader election in population protocols without any kind of oracles. We identify the necessary and sufficient conditions to solve the self-stabilizing leader election in population protocols from the aspects of local memory complexity and fairness assumptions. This paper shows that under the assumption of global fairness, no protocol using only n−1 states can solve the self-stabilizing leader election in complete interaction graphs, where n is the number of agents in the system. To prove this impossibility, we introduce a novel proof technique, called closed-set argument. In addition, we propose a self-stabilizing leader election protocol using n states that works even under the unfairness assumption. This protocol requires the exact knowledge about the number of agents in the system. We also show that such knowledge is necessary to construct any self-stabilizing leader election protocol.