Shoichi Noguchi's research while affiliated with Nihon University and other places

... Models that seek to generate network topologies that are more suited to learning a given problem are classified as adaptive learning algorithms (Anderson & Martinez, 1995; Anderson & Martinez, 1996; Fahlman, & Lebiere, 1990). Partially connected static architectures are also considered in (Chakraborty, Sawada, & Noguchi, 1997). These networks have fewer parameters and are therefore simpler and more efficient than fully connected networks yet are able to perform equally well. ...

... Additionally, the advancement of hardware and software technologies enables neural networks to be applied to exploding NP-complete problems [34]. In one study [35], two neural network models are integrated to solve multiprocessor (parallel resources) deadline problems. Feng et al. applied generic neural networks to solve independent and non-preemptive tasks with deadline requirements for real time scheduling problems [36]. ...

... An attempt to do a NOAA classification with an AI was also done by (Chakraborty et al, 1993). Their research in classification of The NOAA Satellite Image Data by Unsupervised Neural Network was successful. ...

... Nevertheless when there is no reply from the tested node, it is impossible for the tester to determine whether it is the link or the node that has failed or is slow. The test in an arbitrary topology network is thus said to be ambiguous [13]. If a node i sends a test request to node j and times out on the reply, then it is impossible for node i to determine if node j has crashed or alternatively if link ði; jÞ is faulty. ...

... In spite of an abundance of possible applications and an increasing interest in non-Horn logic programming [14] and non-Horn databases [20], extensions of predicate completion (in the sense of a completion taking the form of firstorder clauses) beyond universally reductive predicative Horn clauses are scarce. Reiter [16] and later Togashi et al. [18] proved that predicate completion is possible for non-Horn clauses where every predicate occurs only positively or only negatively and where no two positive atom occurrences in the same clause are unifiable. I showed [9] how to extend predicate completion in a different direction by for the first time allowing non-predicative elements, namely ultimate periodicity equations of the form s n (x)s m (x), which includes reasoning about quotients of the naturals. ...

... As prior work, bistability on S-ALOHA systems has been investigated in [12]- [19]. An analytical framework based on catastrophe theory has been established in [13], where new packet arrival rate and retransmission probabilities (or rate) are identified as two parameters of causing bistability. Based on this framework, the effects of propagation delay and capture, various multipacket reception channels on bistability have been investigated in [14]- [16]. ...

... There are some previous works on the learning of functions and the induction of functional programs, usually combining very dierent techniques (evolutionary programming, MDL principle, folding) to synthesize recursive Standard ML programs as in [24], attening the logic program with function symbols to transform it into an equivalent program without functions [30], using extensions of the least general generalization technique used in ILP (see for instance [1,16,20]) or a similar approach to Shapiros MIS [31] for inducing term rewriting systems [35]. Also, there are some early works on the induction of LISP programs in the seventies (see [33] for a survey), although we do not consider them much related to our approach because they learn from execution traces. ...

... Diagnosis algorithms for arbitrary topology networks, on the other hand, usually rely on periodic tests or heartbeats, which could lead to battery depletion in sensor nodes. This family is represented by algorithms such as [13], [14], [15], [16], [17], [18]. ...

... Untimed bisimulation equivalence is used for checking whether two systems are equivalent in spite of the di erent time constraints. If the corresponding LTS's are nite, we can easily check the two bisimulation equivalences by the algorithms in KS90,SKTN90]. ...

... This is a well-known idea going back at least to [51]. Some more examples of this approach may be found in [3, 39, 40, 48]. By regarding function definitions as syntactical sugar that is transformed away the problem of giving a computational model suitable for both functions and predicates is avoided. ...