arXiv subjects

Atushi Ishikawa

Publications and source records attributed to Atushi Ishikawa.

At least 19 recordsLinked to original sources

arXiv2023

Generating Individual Trajectories Using GPT-2 Trained from Scratch on Encoded Spatiotemporal Data

Following Mizuno, Fujimoto, and Ishikawa's research (Front. Phys. 2022), we transpose geographical coordinates expressed in latitude and longitude into distinctive location tokens that embody positions across varied spatial scales. We encapsulate an individual daily trajectory as a sequence of tokens by adding unique time interval tokens to the location tokens. Using the architecture of an autoregressive language model, GPT-2, this sequence of tokens is trained from scratch, allowing us to construct a deep learning model that sequentially generates an individual daily trajectory. Environmental factors such as meteorological conditions and individual attributes such as gender and age are symbolized by unique special tokens, and by training these tokens and trajectories on the GPT-2 architecture, we can generate trajectories that are influenced by both environmental factors and individual attributes.

cs.LG↗

arXiv2023

Composite distributions in the social sciences: A comparative empirical study of firms' sales distribution for France, Germany, Italy, Japan, South Korea, and Spain

We study 17 different statistical distributions for sizes obtained {}from the classical and recent literature to describe a relevant variable in the social sciences and Economics, namely the firms' sales distribution in six countries over an ample period. We find that the best results are obtained with mixtures of lognormal (LN), loglogistic (LL), and log Student's $t$ (LSt) distributions. The single lognormal, in turn, is strongly not selected. We then find that the whole firm size distribution is better described by a mixture, and there exist subgroups of firms. Depending on the method of measurement, the best fitting distribution cannot be defined by a single one, but as a mixture of at least three distributions or even four or five. We assess a full sample analysis, an in-sample and out-of-sample analysis, and a doubly truncated sample analysis. We also provide the formulation of the preferred models as solutions of the Fokker--Planck or forward Kolmogorov equation.

econ.GN↗

arXiv2010

Shape of Growth Rate Distribution Determines the Type of Non-Gibrat's Property

In this study, the authors examine exhaustive business data on Japanese firms, which cover nearly all companies in the mid- and large-scale ranges in terms of firm size, to reach several key findings on profits/sales distribution and business growth trends. First, detailed balance is observed not only in profits data but also in sales data. Furthermore, the growth-rate distribution of sales has wider tails than the linear growth-rate distribution of profits in log-log scale. On the one hand, in the mid-scale range of profits, the probability of positive growth decreases and the probability of negative growth increases symmetrically as the initial value increases. This is called Non-Gibrat's First Property. On the other hand, in the mid-scale range of sales, the probability of positive growth decreases as the initial value increases, while the probability of negative growth hardly changes. This is called Non-Gibrat's Second Property. Under detailed balance, Non-Gibrat's First and Second Properties are analytically derived from the linear and quadratic growth-rate distributions in log-log scale, respectively. In both cases, the log-normal distribution is inferred from Non-Gibrat's Properties and detailed balance. These analytic results are verified by empirical data. Consequently, this clarifies the notion that the difference in shapes between growth-rate distributions of sales and profits is closely related to the difference between the two Non-Gibrat's Properties in the mid-scale range.

physics.soc-ph↗

arXiv2008

Non-Gibrat's law in the middle scale region

By using numerical simulation, we confirm that Takayasu--Sato--Takayasu (TST) model which leads Pareto's law satisfies the detailed balance under Gibrat's law. In the simulation, we take an exponential tent-shaped function as the growth rate distribution. We also numerically confirm the reflection law equivalent to the equation which gives the Pareto index $μ$ in TST model. Moreover, we extend the model modifying the stochastic coefficient under a Non-Gibrat's law. In this model, the detailed balance is also numerically observed. The resultant pdf is power-law in the large scale Gibrat's law region, and is the log-normal distribution in the middle scale Non-Gibrat's one. These are accurately confirmed in the numerical simulation.

physics.soc-ph↗

arXiv2007

Quasistatically varying log-normal distribution in the middle scale region of Japanese land prices

Employing data on the assessed value of land in 1974--2007 Japan, we exhibit a quasistatically varying log-normal distribution in the middle scale region. In the derivation, a Non-Gibrat's law under the detailed quasi-balance is adopted together with two approximations. The resultant distribution is power-law with the varying exponent in the large scale region and the quasistatic log-normal distribution with the varying standard deviation in the middle scale region. In the distribution, not only the change of the exponent but also the change of the standard deviation depends on the parameter of the detailed quasi-balance. These results are consistently confirmed by the empirical data.

q-fin.ST↗

arXiv2007

The log-normal distribution from Non-Gibrat's law in the middle scale region of profits

Employing profits data of Japanese firms in 2003--2005, we kinematically exhibit the static log-normal distribution in the middle scale region. In the derivation, a Non-Gibrat's law under the detailed balance is adopted together with following two approximations. Firstly, the probability density function of profits growth rate is described as a tent-shaped exponential function. Secondly, the value of the origin of the growth rate distribution divided into bins is constant. The derivation is confirmed in the database consistently. This static procedure is applied to a quasi-static system. We dynamically describe a quasi-static log-normal distribution in the middle scale region. In the derivation, a Non-Gibrat's law under the detailed quasi-balance is adopted together with two approximations confirmed in the static system. The resultant distribution is power-law with varying Pareto index in the large scale region and the quasi-static log-normal distribution in the middle scale region. In the distribution, not only the change of Pareto index but also the change of the variance of the log-normal distribution depends on the parameter of the detailed quasi-balance. As a result, Pareto index and the variance of the log-normal distribution are related to each other.

q-fin.GN↗

arXiv2007

The uniqueness of company size distribution function from tent-shaped growth rate distribution

We report the proof that the extension of Gibrat's law in the middle scale region is unique and the probability distribution function (pdf) is also uniquely derived from the extended Gibrat's law and the law of detailed balance. In the proof, two approximations are employed. The pdf of growth rate is described as tent-shaped exponential functions and the value of the origin of the growth rate distribution is constant. These approximations are confirmed in profits data of Japanese companies 2003 and 2004. The resultant profits pdf fits with the empirical data with high accuracy. This guarantees the validity of the approximations.

physics.soc-ph↗

arXiv2006

The uniqueness of the profits distribution function in the middle scale region

We report the proof that the expression of extended Gibrat's law is unique and the probability distribution function (pdf) is also uniquely derived from the law of detailed balance and the extended Gibrat's law. In the proof, two approximations are employed that the pdf of growth rate is described as tent-shaped exponential functions and that the value of the origin of growth rate is constant. These approximations are confirmed in profits data of Japanese companies 2003 and 2004. The resultant profits pdf fits with the empirical data with high accuracy. This guarantees the validity of the approximations.

physics.soc-ph↗

arXiv2006

Dynamical change of Pareto index in Japanese land prices

We investigate the dynamical behavior in the large scale region of non-equilibrium systems, by employing data on the assessed value of land in 1983 -- 2006 Japan. In the system we find the detailed quasi-balance, which has the symmetry: x_1 -> a {x_2}^θ (x_1 and x_2 are two successive land prices). By using the detailed quasi-balance and Gibrat's law, we derive Pareto's law with varying Pareto index annually. The parameter θcorresponds with the ratio of Pareto indices (μ_1 + 1)/(μ_2 + 1), and the relation is confirmed in the empirical data nicely.

physics.soc-ph↗

arXiv2005

Annual change of Pareto index dynamically deduced from the law of detailed quasi-balance

Employing data on the assessed value of land in 1983 -- 2005 Japan, we investigate the dynamical behavior in the high scale region of non-equilibrium systems. From the detailed quasi-balance and Gibrat's law, we derive a relation between the change of Pareto index and a symmetry in the detailed quasi-balance. The relation is confirmed in the empirical data nicely.

physics.soc-ph↗

arXiv2005

Derivation of the distribution from extended Gibrat's law

Employing profits data of Japanese companies in 2002 and 2003, we identify the non-Gibrat's law which holds in the middle profits region. From the law of detailed balance in all regions, Gibrat's law in the high region and the non-Gibrat's law in the middle region, we kinematically derive the profits distribution function in the high and middle range uniformly. The distribution function accurately fits with empirical data without any fitting parameter.

physics.soc-ph↗

arXiv2005

Pareto index induced from the scale of companies

Employing profits data of Japanese companies in 2002 and 2003, we confirm that Pareto's law and the Pareto index are derived from the law of detailed balance and Gibrat's law. The last two laws are observed beyond the region where Pareto's law holds. By classifying companies into job categories, we find that companies in a small scale job category have more possibilities of growing than those in a large scale job category. This kinematically explains that the Pareto index for the companies in the small scale job class is larger than that for the companies in the large scale job class.

physics.soc-ph↗

arXiv2004

Pareto law and Pareto index in the income distribution of Japanese companies

In order to study the phenomenon in detail that income distribution follows Pareto law, we analyze the database of high income companies in Japan. We find a quantitative relation between the average capital of the companies and the Pareto index. The larger the average capital becomes, the smaller the Pareto index becomes. From this relation, we can possibly explain that the Pareto index of company income distribution hardly changes, while the Pareto index of personal income distribution changes sharply, from a viewpoint of capital (or means). We also find a quantitative relation between the lower bound of capital and the typical scale at which Pareto law breaks. The larger the lower bound of capital becomes, the larger the typical scale becomes. From this result, the reason there is a (no) typical scale at which Pareto law breaks in the income distribution can be understood through (no) constraint, such as the lower bound of capital or means of companies, in the financial system.

cond-mat.other↗

arXiv2004

Relations between a typical scale and averages in the breaking of fractal distribution

We study distributions which have both fractal and non-fractal scale regions by introducing a typical scale into a scale invariant system. As one of models in which distributions follow power law in the large scale region and deviate further from the power law in the smaller scale region, we employ 2-dim quantum gravity modified by the $R^2$ term. As examples of distributions in the real world which have similar property to this model, we consider those of personal income in Japan over latest twenty fiscal years. We find relations between the typical scale and several kinds of averages in this model, and observe that these relations are also valid in recent personal income distributions in Japan with sufficient accuracy. We show the existence of the fiscal years so called bubble term in which the gap has arisen in power law, by observing that the data are away from one of these relations. We confirm, therefore, that the distribution of this model has close similarity to those of personal income. In addition, we can estimate the value of Pareto index and whether a big gap exists in power law by using only these relations. As a result, we point out that the typical scale is an useful concept different from average value and that the distribution function derived in this model is an effective tool to investigate these kinds of distributions.

cond-mat.stat-mech↗

arXiv2003

Fractal Structure with a Typical Scale

In order to understand characteristics common to distributions which have both fractal and non-fractal scale regions in a unified framework, we introduce a concept of typical scale. We employ a model of 2d gravity modified by the $R^2$ term as a tool to understand such distributions through the typical scale. This model is obtained by adding an interaction term with a typical scale to a scale invariant system. A distribution derived in the model provides power law one in the large scale region, but Weibull-like one in the small scale region. As examples of distributions which have both fractal and non-fractal regions, we take those of personal income and citation number of scientific papers. We show that these distributions are fitted fairly well by the distribution curves derived analytically in the $R^2$ 2d gravity model. As a result, we consider that the typical scale is a useful concept to understand various distributions observed in the real world in a unified way. We also point out that the $R^2$ 2d gravity model provides us with an effective tool to read the typical scales of various distributions in a systematic way.

cond-mat.stat-mech↗

arXiv2002

Geometrical Construction of Heterogeneous Loop Amplitudes in 2D Gravity

We study a disk amplitude which has a complicated heterogeneous matter configuration on the boundary in a system of the (3,4) conformal matter coupled to two-dimensional gravity. It is analyzed using the two-matrix chain model in the large N limit. We show that the disk amplitude calculated by Schwinger-Dyson equations can completely be reproduced through purely geometrical consideration. From this result, we speculate that all heterogeneous loop amplitudes can be derived from the geometrical consideration and the consistency among relevant amplitudes.

hep-th↗

arXiv2002

A New Approach to Personal Income Distribution

The results of R^2 dynamical random surface model (2-dimensional quantum gravity with a $R^2$ term) are applied to explain the personal income distribution. A scale invariance exists if there is not the $R^2$ term in the action. The R^2 term provides a typical scale and breaks the scale invariance explicitly in the low and middle income range. A new distribution, Weibull distribution, is deduced from the action analytically in the low income range, and a consistent fitting is obtained in the whole income range. Also, we show that the lognormal distribution in the middle income range can be understood in this framework.

cond-mat.stat-mech↗

arXiv2000

Splitting of Heterogeneous Boundaries in a System of the Tricritical Ising Model Coupled to 2-Dim Gravity

We study disk amplitudes whose boundaries have heterogeneous matter states in a system of $(4,5)$ conformal matter coupled to 2-dim gravity. They are analysed by using the 3-matrix chain model in the large $N$ limit. Each of the boundaries is composed of two or three parts with distinct matter states. From the obtained amplitudes, it turns out that each heterogeneous boundary loop splits into several loops and we can observe properties in the splitting phenomena that are common to each of them. We also discuss the relation to boundary operators.

hep-th↗