This paper investigates the characteristics of the associative memories of Hopfield networks(HF) and cellular neural networks(CNN). In a Hopfield network, the memories which are to be stored correspond to the extrema of the energy function defined in the network, and in a CNN, they correspond to the asymptotically stable points of the differential equation of the CNN. The associative memory networks of Chinese characters are simulated on a 9x9 matrix network. And it is found that the memory capacity of a CNN is larger than that of an HF by numerical simulations. In a 3x3 matrix network, the characteristics of the associative memories of Hopfield networks and cellular neural networks are examined in details.
keywords: Interconnecting networkA procedure to interpret principal component axes is proposed by making use of hypothesis-testing procedure on the means of observed variables between the positive and negative sign groups which are defined by the observed principal component scores. It gives similar analytic properties as the usual procedure based on the eigen vectors of the covariance matrix, when the observed vector x conforms to a multivariate normal distribution. It is shown to work well by two examples from Mardia, Kent & Bibby(1979) and Tanaka, Tarumi & Wakimoto(1984).
keyword: Test on means, Normal theoryRedundancy analysis(RDA) was developed originally for descriptive data analysis of two sets of quantitative (continuous) variables, one set being dependent on the other. Isra"els(1984) generalized it to the case of qualitative (categorical) variables. In this paper we suggest a procedure(QRDA), extending RDA to deal with data described in terms of both qualitative and quantitative variables. For its evaluation, we give a Monte Carlo study to examine the stability of data analysis by QRDA, and demonstrate its validity . Further we discuss properties of QRDA in comparison with Isra"els' procedure on the point of eigenvalue problems, stating that the two procedures give the same eigenvector under a special condition on the data matrix. Then we show, on the basis of simulation study, that QRDA is more advantageous than Isra"els' procedure even in analyzing data in terms of qualitative variables only.
keywords: Descriptive data analysis, Monte Carlo study, Qualitative data, Quantitative data
The Internet is claimed to be very democratic medium. Every person
can transmit and receive wide variety of information at the very low cost.
But this causes the ``flood'' of problematic expression including
pornography and hate speech as well.
This article first sketches the various efforts to control these
information by countries and an international organization including the
United States, the European Union, France, Germany, the United Kingdom,
Singapore and Japan.
The United States has chosen to suppress ``indecent'' expression on line by
law, namely ``Communications Decency Act of 1996.'' The constitutionality of
the Act was immediately challenged in courts. Two of the federal district
courts ruled the law unconstitutional, and the Supreme Court is expected to
reveal its opinion by this summer.
In contrast Japanese Ministry of Telecommunications is planning to
avoid legal restriction and to give private Internet-related organizations
the primary responsibility for settling the complaints from the Internet users.
In addition, the Ministry is encouraging to develop practical filtering
software that screens content that is harmful to minors. It is afraid,
however, that the government might exert substantial control over the
problem in terms of providing ``guidelines'' that ``lacks'' any legal effect
but has de facto binding power in a society like Japanese.
The article concludes by pointing out the doctrinal issues
underlying the content restriction in the cyberspace. (1) In a society
where all of the effective communicative tools are virtually monopolized by
the mass media, the Internet has superior constitutional value considering
its highly democratic character. (2) The protection of minors provides good
reason to restrict harmful content in theory, but this type of restriction
usually turns out to be unconstitutionally overbroad constraint in
practice. The legal protection of children may become unnecessary with
installation of proper filtering mechanism. (3) As suggested above,
``voluntary restraint'' may have strong restrictive power due to the
nature of the society, and warrants close constitutional analysis.