Publisher Summary This chapter provides an account of different neural network architectures for pattern recognition. A neural network consists of several simple processing elements called neurons. Each neuron is connected to some other neurons and possibly to the input nodes. Neural networks provide a simple computing paradigm to perform complex recognition tasks in real time. The chapter categorizes neural networks into three types: single-layer networks, multilayer feedforward networks, and feedback networks. It discusses the gradient descent and the relaxation method as the two underlying mathematical themes for deriving learning algorithms. A lot of research activity is centered on learning algorithms because of their fundamental importance in neural networks. The chapter discusses two important directions of research to improve learning algorithms: the dynamic node generation, which is used by the cascade correlation algorithm; and designing learning algorithms where the choice of parameters is not an issue. It closes with the discussion of performance and implementation issues.

Neural Networks for Pattern Recognition

artificial neural networks , artificial neural networks , مرکز فناوری اطلاعات و اطلاع رسانی کشاورزی

Artificial Neural Networks

Fuzzy cognitive maps FCM can structure virtual worlds that change with time. An FCM links causal events, actors, values, goals, and trends in a fuzzy feedback dynamical system. An FCM lists the fuzzy rules or causal flow paths that relate events. It can guide actors in a virtual world as the actors move through a web of cause and effect and react to events and to other actors. Experts draw FCM causal pictures of the virtual world. They do not write down differential equations to change the virtual world. Complex FCMs can give virtual worlds with “new” or chaotic equilibrium behavior. Simple FCMs give virtual worlds with periodic behavior. They map input states to limit-cycle equilibria. An FCM limit cycle repeats a sequence of events or a chain of actions and responses. Limit cycles can control the steady-state rhythms and patterns in a virtual world. In nested FCMs each causal concept can control its own FCM or fuzzy function approximator. This gives levels of fuzzy systems that can choose goals and causal webs as well as move objects and guide actors in the webs. FCM matrices sum to give a combined FCM virtual world for any number of knowledge sources. Adaptive FCMs change their fuzzy causal web as causal patterns change and as actors act and experts state their causal knowledge. Neural learning laws change the causal rules and the limit cycles. Actors learn new patterns and reinforce old ones. In complex FCMs the user can choose the dynamical structure of the virtual world from a spectrum that ranges from mildly to wildly nonlinear. We use a simple but adaptive FCM to model an undersea virtual world of dolphins, fish, and sharks.

Virtual worlds as fuzzy cognitive maps

Morphological bidirectional associative memories

https://ink.library.smu.edu.sg/cgi/viewcontent.cgi?article=6227&context=sis_research

Adaptive resonance associative map

Enhancements of the encoding strategy of a discrete bidirectional associative memory (BAM) reported by B. Kosko (1987) are presented. There are two major concepts in this work: multiple training, which can be guaranteed to achieve recall of a single trained pair under suitable initial conditions of data, and dummy augmentation, which can be guaranteed to achieve recall of all trained pairs if attaching dummy data to the training pairs is allowable. In representative computer simulations, multiple training has been shown to lead to an improvement over the original Kosko strategy for recall of multiple pairs as well. A sufficient condition for a correlation matrix to make the energies of the training pairs be local minima is discussed. The use of multiple training and dummy augmentation concepts are illustrated, and theorems underlying the results are presented. >

Two coding strategies for bidirectional associative memory

Necessary and sufficient conditions are derived for the weights of a generalized correlation matrix of a bidirectional associative memory (BAM) which guarantee the recall of all training pairs. A linear programming/multiple training (LP/MT) method that determines weights which satisfy the conditions when a solution is feasible is presented. The sequential multiple training (SMT) method is shown to yield integers for the weights, which are multiplicities of the training pairs. Computer simulation results, including capacity comparisons of BAM, LP/MT BAM, and SMT BAM, are presented. >

Guaranteed recall of all training pairs for bidirectional associative memory

The minimal number of times for using a pair for training to guarantee recall of that pair among a set of training pairs is derived for a bidirectional associative memory. >

On multiple training for bidirectional associative memory

Multiple training concept for back-propagation neural networks for use in associative memories

Adaptive scheduling for flexible manufacturing systems using a neural network structure is described. Adaptive scheduling consists of three parts: features representation of the job orders and the factory environment, a set of fixed scheduling algorithms, and an algorithm selector. The statistics on orders is represented by a small number of feature parameters. Each of the algorithms included in the algorithm set is most suitable for a specific set of orders and factory status, but no single algorithm is best under all conditions. The algorithm selector chooses one of the algorithms to optimize some scheduling performance measure, for given values of the feature parameters. To illustrate the performance of the adaptive scheduling process, 10,000 different order cases were taken from uniformly distributed data. For the simulation, the algorithm selector was implemented using a backpropagation neural network structure. The results of the computer experiments demonstrated the superiority of adaptive scheduling over all of the seven fixed scheduling strategies. >

Y.-F. Wang

Papers

Two coding strategies for bidirectional associative memory

Guaranteed recall of all training pairs for bidirectional associative memory

On multiple training for bidirectional associative memory

Multiple training concept for back-propagation neural networks for use in associative memories

Adaptive scheduling utilizing a neural network structure