Motonobu Hattori

Chaotic bidirectional associative memory

A chaotic bidirectional associative memory (CBAM) is proposed and simulated. It can deal with one-to-many associations such as {(A,a), (A,b), (A,c),/spl middot//spl middot//spl middot/}. In the CBAM, in order to enable one-to-many associations, each training pair is memorized together with its own contextual information and chaotic neurons are employed in a part of the network corresponding to the contextual information. Since the chaotic neurons change their states by chaos, the one-to-many associations can be realized in the CBAM.

Improved multidirectional associative memories for training sets including common terms

Improved multidirectional associative memories (IMAMs) are proposed and simulated. The IMAM fundamental component is a multilayer neural network. IMAMs can memorize and recall multiple associations even when training sets include common terms, such as the training sets composed of (A,a,1), (A,b,2), (C,b,3). The structure of the proposed IMAMs is represented by mutual connections of multilayer neural networks. The proposed IMAMs require less parameters compared with other associative memories and are capable of automatic recall. Recall performance can be greatly improved by using a priority coefficient.<<ETX>>

Episodic associative memories

Abstract In this paper, we propose Episodic Associative Memories (EAMs). They use Quick Learning for Bidirectional Associative Memory (QLBAM), which enables high memory capacity, and Pseudo-Noise (PN) sequences. In the learning of the proposed EAMs, PN sequences are used as one side of training pairs. To store an episode, a scene of the episode is stored with a PN sequence, and the next scene is stored with the PN sequence which is shifted with one bit. Such a procedure enables episodic memory. The proposed EAMs can recall episodic associations by shifting the PN sequence one by one. The features of the proposed EAMs are: (1) they can memorize and recall episodic associations; (2) they can store plural episodes; (3) they have high memory capacity; (4) they are robust for noisy and incomplete inputs.

Associative memory for intelligent control

In many industrial applications of softcomputing, intelligent controls are important to accomplish high level tasks. Intelligent controls, however, need specific knowledge for each task. Therefore developing good memory is crucial to store the required knowledge efficiently and robustly. Neural network associative memories are the most suitable for the role because of their flexibility and content addressability. In this paper, first, we describe the basic concept of the neural network associative memories and the conventional learning algorithms. After pointing out some problems of the associative memories, we explain a novel learning algorithm, which is superior to the conventional ones. Finally, we introduce an associative memory suited for the intelligent controls and show the effectiveness by a number of computer simulations.

Quick learning for multidirectional associative memories

In this paper, a quick learning algorithm for multidirectional associative memories (MAMs) is proposed. With this quick learning algorithm, not only the storage capacity of the MAMs can be improved, but also the recall of all training data can be guaranteed. In addition, several important characteristics of the MAMs such as the relation between the required learning epochs and the number of layers, and the relation between the noise reduction effect and the number of layers are introduced.

Chaotic multidirectional associative memory

A chaotic multidirectional associative memory (CMAM) is proposed and simulated. It can deal with many-to-many associations and the structure is very simple. Furthermore, similarity to a psychological fact (priming effect) is observed in the association of the CMAM. In order to enable many-to-many associations, the CMAM memorizes each training data together with its own contextual information and employs chaotic neurons. Since the chaotic neurons change their states by chaos, many-to-many associations can be realized in the CMAM.

Episodic associative memory

Episodic associative memory (EAM) is introduced and simulated. It uses quick learning for bidirectional associative memory (QLBAM) and pseudo-noise (PN) sequences. The features of the proposed EAM are: it can memorize and recall episodic associations; it can store plural episodes; it has high memory capacity.<<ETX>>

Knowledge processing system using multidirectional associative memory

A knowledge processing system using multidirectional associative memory (KP-MAM) is proposed and simulated. The KP-MAM uses distributed knowledge representation and can represent knowledge stored in a form of semantic network. In order to realize knowledge processing system using distributed representation, the KP-MAM enables the following five functions: associations of three-tuple relation, characteristics inheritance, exception processing, many-to-many associations and high storage capacity. The proposed KP-MAM has the following features: (1) It can realize knowledge processing using distributed representation; (2) It is robust for noisy inputs; (3) It guarantees recall of all training data owing to a quick learning algorithm; (4) It has high storage capacity.

Intersection learning for bidirectional associative memory

We propose intersection learning for bidirectional associative memory (ILBAM). ILBAM is based on a novel relaxation method. A number of computer simulations show the effectiveness of ILBAM: (1) it can guarantee the recall of all training pairs; (2) it requires much lower weight renewal times than conventional methods; (3) it becomes more effective in the case where there are many training pairs needed to be stored; (4) it is insensitive to the correlation of training pairs; and (5) it contributes to the noise reduction effect of BAM.

Multimodule associative memory for many-to-many associations

A multimodule associative memory for many-to-many associations (MMA)/sup 2/ is proposed. The features of the proposed (MMA)/sup 2/ are: 1) it can memorize and recall not only many-to-many associations but also the context and union associations; 2) it can guarantee the recall of all training data; and 3) it has high storage capacity.