arXiv subjects

Kazuki Fukui

Publications and source records attributed to Kazuki Fukui.

3 recordsLinked to original sources

arXiv2019

Integration of Liquid Thermoelectrochemical Conversion into Forced Convection Cooling

Forced convection cooling is important in numerous technologies ranging from microprocessors in data centers to turbines and engines; active cooling is essential in these situations. However, active transfer of heat or thermal energy under a large temperature difference promptly destroys the exergy, which is the free-energy component of thermal energy, and this issue has remained unaddressed. Herein, we describe a thermoelectrochemical conversion to partially recover presently lost exergy in forced convection cooling. We design a test cell in which an electrolyte liquid is forced through a channel formed between two parallel electrodes and the hot-side electrode simulates an object to be cooled. Our investigations show that the narrower interelectrode channels afford higher cooling and power generation performances. The mass transfer resistance is the most dominant type of resistance for all the conditions tested and the charge transfer kinetics is likely to be controlled by viscosity. The dependence of the generated power on the flow rate is caused by the change in the diffusion coefficient of redox species with temperature. As an evaluation measure for such forced-flow thermocells, the gain (Λ) --defined as the ratio of the generated power to the hydrodynamic pumping work required to force the liquid through the cell-- is introduced. Λis above unity in a certain flow rate region. This demonstrates that such a system can generate more electric power than the pump work required to drive the liquid through the cell, suggesting its potential to partly recover presently lost exergy of thermal energy as electricity.

physics.chem-ph↗

arXiv2019

Representation Learning with Weighted Inner Product for Universal Approximation of General Similarities

We propose $\textit{weighted inner product similarity}$ (WIPS) for neural network-based graph embedding. In addition to the parameters of neural networks, we optimize the weights of the inner product by allowing positive and negative values. Despite its simplicity, WIPS can approximate arbitrary general similarities including positive definite, conditionally positive definite, and indefinite kernels. WIPS is free from similarity model selection, since it can learn any similarity models such as cosine similarity, negative Poincaré distance and negative Wasserstein distance. Our experiments show that the proposed method can learn high-quality distributed representations of nodes from real datasets, leading to an accurate approximation of similarities as well as high performance in inductive tasks.

cs.LG↗

arXiv2019

Segmentation-free Compositional $n$-gram Embedding

We propose a new type of representation learning method that models words, phrases and sentences seamlessly. Our method does not depend on word segmentation and any human-annotated resources (e.g., word dictionaries), yet it is very effective for noisy corpora written in unsegmented languages such as Chinese and Japanese. The main idea of our method is to ignore word boundaries completely (i.e., segmentation-free), and construct representations for all character $n$-grams in a raw corpus with embeddings of compositional sub-$n$-grams. Although the idea is simple, our experiments on various benchmarks and real-world datasets show the efficacy of our proposal.

cs.CL↗