Associative memory for online learning in noisy environments using self-organizing incremental neural network

Associative memory operating in a real environment must perform well in online incremental learning and be robust to noisy data because noisy associative patterns are presented sequentially in a real environment. We propose a novel associative memory that satisfies these requirements. Using the proposed method, new associative pairs that are presented sequentially can be learned accurately without forgetting previously learned patterns. The memory size of the proposed method increases adaptively with learning patterns. Therefore, it suffers neither redundancy nor insufficiency of memory size, even in an environment in which the maximum number of associative pairs to be presented is unknown before learning. Noisy inputs in real environments are classifiable into two types: noise-added original patterns and faultily presented random patterns. The proposed method deals with two types of noise. To our knowledge, no conventional associative memory addresses noise of both types. The proposed associative memory performs as a bidirectional one-to-many or many-to-one associative memory and deals not only with bipolar data, but also with real-valued data. Results demonstrate that the proposed method's features are important for application to an intelligent robot operating in a real environment. The originality of our work consists of two points: employing a growing self-organizing network for an associative memory, and discussing what features are necessary for an associative memory for an intelligent robot and proposing an associative memory that satisfies those requirements.

A feedforward bidirectional associative memory

In contrast to conventional feedback bidirectional associative Memory (BAM) network models, a feedforward BAM network is developed based on a one-shot design algorithm of O(p(2)(n+m)) computational complexity, where p is the number of prototype pairs and n, m are the dimensions of the input/output bipolar vectors. The feedforward BAM is an n-p-m three-layer network of McCulloch-Pitts neurons with storage capacity 2(min{m,n}) and guaranteed perfect bidirectional recall. The overall network design procedure is fully scalable in the sense that any number p= or <2(min{m,n}) of bidirectional associations can be implemented. The prototype patterns may be arbitrarily correlated. With respect to inference performance, it is shown that the Hamming attractive radius of each prototype reaches the maximum possible value. Simulation studies and comparisons illustrate and support these theoretical developments.

A bidirectional heteroassociative memory for binary and grey-level patterns

Typical bidirectional associative memories (BAM) use an offline, one-shot learning rule, have poor memory storage capacity, are sensitive to noise, and are subject to spurious steady states during recall. Recent work on BAM has improved network performance in relation to noisy recall and the number of spurious attractors, but at the cost of an increase in BAM complexity. In all cases, the networks can only recall bipolar stimuli and, thus, are of limited use for grey-level pattern recall. In this paper, we introduce a new bidirectional heteroassociative memory model that uses a simple self-convergent iterative learning rule and a new nonlinear output function. As a result, the model can learn online without being subject to overlearning. Our simulation results show that this new model causes fewer spurious attractors when compared to others popular BAM networks, for a comparable performance in terms of tolerance to noise and storage capacity. In addition, the novel output function enables it to learn and recall grey-level patterns in a bidirectional way.

Sensitivity to noise in bidirectional associative memory (BAM)

Original Hebbian encoding scheme of bidirectional associative memory (BAM) provides a poor pattern capacity and recall performance. Based on Rosenblatt's perceptron learning algorithm, the pattern capacity of BAM is enlarged, and perfect recall of all training pattern pairs is guaranteed. However, these methods put their emphases on pattern capacity, rather than error correction capability which is another critical point of BAM. This paper analyzes the sensitivity to noise in BAM and obtains an interesting idea to improve noise immunity of BAM. Some researchers have found that the noise sensitivity of BAM relates to the minimum absolute value of net inputs (MAV). However, in this paper, the analysis on failure association shows that it is related not only to MAV but also to the variance of weights associated with synapse connections. In fact, it is a positive monotone increasing function of the quotient of MAV divided by the variance of weights. This idea provides an useful principle of improving error correction capability of BAM. Some revised encoding schemes, such as small variance learning for BAM (SVBAM), evolutionary pseudorelaxation learning for BAM (EPRLAB) and evolutionary bidirectional learning (EBL), have been introduced to illustrate the performance of this principle. All these methods perform better than their original versions in noise immunity. Moreover, these methods have no negative effect on the pattern capacity of BAM. The convergence of these methods is also discussed in this paper. If there exist solutions, EPRLAB and EBL always converge to a global optimal solution in the senses of both pattern capacity and noise immunity. However, the convergence of SVBAM may be affected by a preset function.

On solving systems of linear inequalities with artificial neural networks

The implementation of the relaxation-projection algorithm by artificial neural networks to solve sets of linear inequalities is examined. The different versions of this algorithm are described, and theoretical convergence results are given. The best known analog optimization solvers are shown to use the simultaneous projection version of it. Neural networks that implement each version are described. The results of tests, made with simulated realizations of these networks, are reported. These tests consisted in having all networks solve some sample problems. The results obtained help determine good values for the step size parameters, and point out the relative merits of the different networks.