In this paper, several possible algorithms and training methods for the CMAC network are analyzed thoroughly. Improvements are then examined and a Hybrid approach has been developed for the Maximum Error Algorithm by using the Neighborhood Training for the initial training period. The employment of the technique yielded faster initial convergence which is very important for many control applica...

We found that the training process results declined when the average number of missing pixels in the training flow maps was too high. Some of the matching algorithms, in particular DeepMatching, did produce sparse maps like these. To tackle this problem, we calculate the flow map bi-directionally (From I to I ′ and from I ′ to I) using the matching algorithm. We invert the second flow map and a...

Due to its high spectrum efficiency and good ability in coping with multi-path fading, orthogonal frequency division multiplexing (OFDM) technology has attracted considerable attention from researchers. However, OFDM system is very sensitive to symbol timing offset and carrier frequency offset, so accurate timing synchronization and frequency synchronization are necessary to OFDM system. This p...

We propose a novel strategy for training neural networks using sequential Monte Carlo algorithms. This global optimisation strategy allows us to learn the probability distribution of the network weights in a sequential framework. It is well suited to applications involving on-line, nonlinear or non-stationary signal processing. We show how the new algorithms can outperform extended Kalman lter ...

Architectures of neural networks affect the training performance artificial networks. For more consistent evaluation algorithms, hard-to-train benchmarking architectures should be used. This study introduces a benchmark network architecture, which is called pipe-like and presents analyses for popular Neural Network Backpropagation Algorithms (NNBA) well-known Metaheuristic Search (MSA). The ess...

Many training algorithms (like gradient descent, for example) use random initial weights. These algorithms are rather sensitive to their starting position in the error space, which is represented by their initial weights. This paper shows that the training performance can be improved signiicantly by using a Particle Swarm Optimizer (PSO) to initialize the weights, rather than random initializat...

One important subtask of Referring Expression Generation (REG) algorithms is to select the attributes in a definite description for a given object. In this paper, we study how much training data is required for algorithms to do this properly. We compare two REG algorithms in terms of their performance: the classic Incremental Algorithm and the more recent Graph algorithm. Both rely on a notion ...

Metonymy recognition is generally approached with complex algorithms that rely heavily on the manual annotation of training and test data. This paper will relieve this complexity in two ways. First, it will show that the results of the current learning algorithms can be replicated by the ‘lazy’ algorithm of Memory-Based Learning. This approach simply stores all training instances to its memory ...