Training the neocognitron network using design of experiments

doi:10.1016/0954-1810(95)95752-R

Artificial Intelligence in Engineering

Volume 9, Issue 2, 1995, Pages 85-94

https://doi.org/10.1016/0954-1810(95)95752-R Get rights and content

Abstract

Neocognitron is a hierarchically structured multi-layer neural network that is recognised for its ability to tolerate scaling and rotational effects of the input patterns. However, there are many parameters that describe the hierarchical multi-layer structure of the network which must be adjusted properly for the network to recognise a set of patterns to allow for their scaling and rotational effects.

The methodology utilises an orthogonal array to determine combination of values of the various parameters, instead of all the possible combinations, thus greatly reducing the number of trials required to arrive at the optimum values.

References (7)

K. Fukushima et al.
Neocognitron: a new algorithm for pattern recognition tolerant of deformations and shifts in position
Pattern Recognition
(1982)
K. Fukushima
Neocognitron: a hierarchical neural network capable of visual pattern recognition
Neural Networks
(1988)
P.D. Wasserman

There are more references available in the full text version of this article.

Cited by (13)

Algorithm architectures for patient dependent seizure detection
2006, Clinical Neurophysiology
The goal of this work is to determine whether improved performance (compared to patient independent algorithms) can be achieved by an algorithm, developed on the fly, that requires no user input beyond the identification of the first one or two seizures in the record.
The previously developed AutoLearn™ algorithm, which employs the probabilistic neural network (PNN), is tested on 209 seizures obtained from the epilepsy monitoring unit (EMU) or ambulatory recordings. A construction algorithm is used to compare a variety of algorithm architectures and factors. The Taguchi design of experiments (DoE) method is employed find the significant factors without resorting to a full factorial design.
Architectures that train a single PNN per channel and use segmentation to identify ranges of similar activity are preferred. The two best architectures are insensitive to the levels of any of the other factors tested. The training time for the algorithm is less than 1 s, and approximately 2 min are required to find the seizures in an 8 h record.
The final algorithm, which requires no input from the user other than the marking of the first seizure in a record, performs as well or better than the 3 seizure detectors on EMU and ambulatory records. The algorithm performs nearly as well as human experts on the EMU records.
The described method can be used to identify unusual seizures (or other patterns) that will be missed by the current generation of seizure detectors. We expect that the methods developed here will also aid the development of patient independent seizure detectors that can improve their performance over time by incorporating new examples.
Robust design of multilayer feedforward neural networks: An experimental approach
2004, Engineering Applications of Artificial Intelligence
Citation Excerpt :
In passing, it is worth noting that the robust design methodology has been applied for optimizing other types of neural networks. For instance, Teo and Sim (1995), Lin and Chen (1996), and Lin and Tseng (2000) adopted this methodology for designing a neocognitron, a fuzzy-net, and a learning vector quantization net, respectively. To the best of the authors’ knowledge, no attempt has been made to apply the Taguchi robust design methodology for optimally designing unsupervised ANNs except Teo and Sim (1995) who deal with a neocognitron as mentioned earlier.
Artificial neural networks (ANNs) have been successfully used for solving a wide variety of problems. However, determining a suitable set of structural and learning parameter values for an ANN still remains a difficult task. This article is concerned with the robust design of multilayer feedforward neural networks trained by backpropagation algorithm (Backpropagation net, BPN) and develops a systematic, experimental strategy which emphasizes simultaneous optimization of BPN parameters under various noise conditions. Unlike previous works, the present robust design problem is formulated as a Taguchi's dynamic parameter design problem, together with a fine-tuning of the BPN output when necessary. A series of computational experiments are also conducted using the data sets from various sources. From the computational results, statistically significant effects of the BPN parameters on the robustness measure (i.e., signal-to-noise ratio) are identified, based upon which an economical experimental strategy is derived. It is also shown that fine-tuning the BPN output is effective in improving the signal-to-noise ratio. Finally, the step-by-step procedures for implementing the proposed approach are illustrated with an example.
Optimum design for Artificial Neural Networks: An example in a bicycle derailleur system
2000, Engineering Applications of Artificial Intelligence
The integration of neural networks and optimization provides a tool for designing network parameters and improving network performance. In this paper, the Taguchi method and the Design of Experiment (DOE) methodology are used to optimize network parameters. The users have to recognize the application problems and choose a suitable Artificial Neural Network model. Optimization problems can then be defined according to the model. The Taguchi method is first applied to a problem to find out the more important factors, then the DOE methodology is used for further analysis and forecasting. A Learning Vector Quantization example is shown for an application to bicycle derailleur systems.
A prototype fuzzy-based classifier system for trajectory planning
1997, Journal of Network and Computer Applications
This paper describes a prototype fuzzy-based classifier system for manipulator trajectory planning. The prototype system comprises four modules: a vision system, an object location analyser, a classifier system and a translator. The vision system is employed to capture and process the image of an object positioned in the work envelope of the manipulator. The object location analyser, which consists of a two-level fuzzy inferencing mechanism, is used to define the relative position of an object with respect to the end effector position. With the position defined and used as the input message from the environment, the classifer system is then employed to plan the trajectory. The translator provides the facility to convert the trajectory obtained into manipulator instructions which are used for the ultimate control of the manipulator. Details of the prototype system are presented. Validation of the prototype system was carried out. Details of the results obtained are discussed.
Artificial Neural Networks (ANNs) as a Novel Modeling Technique in Tribology
2019, Frontiers in Mechanical Engineering
Multidimensional Hebbian Learning with Temporal Coding in Neocognitron Visual Recognition
2017, IEEE Transactions on Systems, Man, and Cybernetics: Systems

View all citing articles on Scopus

View full text

Training the neocognitron network using design of experiments

Abstract

Pattern Recognition

Neural Networks