Shohachiro Nakanishi

Hidden information maximization for feature detection and rule discovery

In this paper, We propose a method to maximize the hidden information stored in hidden units. The hidden information is defined by the decrease in uncertainty of hidden units with respect to input patterns. By maximizing the hidden information, the hidden unit can detect features and extract rules behind input patterns. Our method was applied to two problems: an autoencoder to produce six alphabet letters and the assimilation for the formation of plurals and nasalization in an artificial language. In the first problem, the results explicitly confirmed that the features of input patterns could be detected by maximizing the hidden information. In the second experiment, we could clearly see that the rules of the assimilation were extracted by maximizing the hidden information, even if the rules are obscured by some other factors.

Distributed auction servers resolving winner and winning bid without revealing privacy of bids

We have developed an electronic auction system with a set of distributed servers that collaborate to resolve a winning bid without revealing the secrecy of bids. This paper describes the system architecture and shows the performance in terms of the round complexity and the bandwidth consumption for a message. Based on the actual measurements, the upper bound of number of bidding prices is identified. In addition, by adding some steps to the basic auction protocol in Kikuchi, Harkavy, and Tygar, 1998, we improve the security of protocol so that the second highest bid must not be known even by the winner.

Performance evaluation of public-key certificate revocation system with balanced hash tree

A new method for updating certificate revocation trees (CRT) is proposed. Efficient revocation of public-key certificates is a current issue in public-key infrastructure because a traditional certificate revocation list uses a large amount of bandwidth. A certificate revocation tree is a hash tree of revoiced certificates and reduces a bandwidth consumption up to O(log(n)). In this paper, an implementation of certificate revocation tree with S-expression is presented and the performance of the system is evaluated in terms of communication and computational costs. To update a CRT, we have two algorithms; (1) random insertion-a new certificate to be revoiced is just inserted into the existing tree and (2) balancing updating-balances CRT every time a new certificate is added.

Performance Evaluation of Certificate Revocation Using k-Valued Hash Tree

A CRL (Certificate Revocation List) defined in X.509 is currently used for certificate revocation. There are some issues of CRL including a high communication cost and a low latency for update. To solve the issues, there are many proposals including CRT (Certificate Revocation Tree), Authenticated Dictionary, and Delta List. In this paper, we study CRT using k-valued hash tree. To estimate the optimal value of k, we examine the overhead of computation and the communication cost. We also discuss when a CRT should be reduced by eliminating unnecessary entries that are already expired.

Color planning by fuzzy set theory

The authors propose a system to plan a color arrangement in interior space such as rooms and offices. The system is based on three main processes: a determination process of dominant colors, a determination process of harmony colors, and an adjustment process of the resultant colors. In the first process, the dominant color of the interior space is determined from the linguistic representation of the users image for this space by fuzzy set theory. In the second process, the harmony colors for the dominant color are calculated by P. Moon and D. E. Spencer-Spencers harmony theory, taking into account the area effect and linguistic inputs. If users cannot satisfy the resultant colors represented on a CRT display, the system can adjust the resultant colors to the users favorite color by linguistic representation in the final process.<<ETX>>

Improving generalization performance by information minimization

In this paper, we attempt to show that the information stored in networks must be as small as possible for the improvement of the generalization performance under the condition that the networks can produce targets with appropriate accuracy. The information is defined by the difference between maximum entropy or uncertainty and observed entropy. Borrowing a definition of fuzzy entropy, the uncertainty function is defined for the internal representation and represented by the equation: -/spl upsi//sub i/ log /spl upsi//sub i/-(1-/spl upsi//sub i/) log (1-/spl upsi//sub i/), where /spl upsi//sub i/ is a hidden unit activity. After having formulated an update rule for the minimization of the information, we applied the method to a problem of language acquisition: the inference of the past tense forms of regular verbs. Experimental results confirmed that by our method, the information was significantly decreased and the generalization performance was greatly improved.<<ETX>>

Web Content Transducing System for Cellular Phones

With the proliferation of Internet-accessible cellular phones, Web browsing via cellular phones has increased in recent years. However, the amount of information acquired by using cellular phones is considerably lesser than that acquired by using personal computers because the Web pages created for personal computers are not suited for browsing via cellular phones. In this paper, we implement an automatic Web content transducing system for cellular phones.

Uniqueness of partially specified multiple-valued Kleenean function

A multiple-valued Kleenean function is a mapping f/sub k/:[0,1]/sup n//spl rarr/[0,1], which is representable by a logic formula consisting n variables, three logical connectives, and any constant value of [0,1]. In this paper, some properties of the partially specified multiple-valued Kleenean functions by a subset A of [0,1] are investigated and the identification problem of logic formula is solved. The main result is a necessary and sufficient condition that a mapping f:A/spl rarr/[0,1] to be a unique Kleenean function.

Structure Identification of Acquired Knowledge in Fuzzy Inference by Genetic Algorithms

In the fuzzy modelling and construction of fuzzy inference rules for fuzzy controllers, it is a very important problem to acquire automatically the knowledge for the objects from only their given data. Many methods for knowledge acquisition have been reported and published in the many journals and proceedings in the conference. However, there are no method to identify the structure of acquired knowledge. Then the authors propose a method to identify the structure of the acquired knowledge for objective systems in the form of the multi-stage fuzzy inference from only their given input and output data by a genetic algorithm

Fuzzy system identification for composite operation and fuzzy relation by genetic algorithms

Genetic Algorithms (GA) are a useful and convenient tool to find the solution in combinatorial optimal problems, and widely used in the various engineering fields. Here we apply GA to identify both of the composite operations and fuzzy relations under that operation at the same time from the given input-output system data. There exist many composite operations and associated fuzzy relations, which satisfy the same input-output system data. Then, it is supposed that many composite operations and fuzzy relations, which satisfy the original data, are generated when we apply GA to this problems. Tne authors propose a method to identify the fuzzy system from these composite operations and fuzzy relations, generated by GA, by an unweighted pair-group method using arithmetic average (UPGMA) which was developed to make a taxonomic tree of the expression in molecular biology.