A binary Self-Organizing Map and its FPGA implementation

Abstract

The vector components of the winning node Wk with minimum distance Di; is then updated as follows where TJ is the learning rate. The topological ordering property is imposed by also updating weight vectors of nodes in the neighbourhood of the winning node. This can be achieved by the following learning rulewhere N j is a neighbourhood function (defining the region around Wk ) based on the topological displacement of neighbouring neuron from the winning neuron. The size of N j decreases as training progresses.In the vast majority of implementations, the SOM input data and neurons are represented by real numbers, making it difficult to implement on a hardware architecture like the Field Programmable Gate Array (FPGA). However, in many applications the data is either presented as a binary string, or may be conveniently recoded as such (a "binary signature"). For example, in image processing applications a bank of Haar filters produces a long binary signature. In this paper we present a new learning algorithm which takes binary inputs and maintains tri-state weights (neuron) in the SOM. We also present the FPGA implementation of this binary Self Organizing Map (bSOM). The bSOM is designed for efficient hardware implementation, having both greatly reduced circuit size compared to a real-valued SOM, and exceptionally fast execution and training times.In section II, we review previous implementations of SOM on hardware architectures. The novel bSOM algorithm is then presented in III, followed by its FPGA implementation in section IV. Section V, presents the experimental results in software and hardware, and we conclude in section VI.During training, the "nearest" neuron prototype vector to the input vector is identified -this is called the "winning" neuron -using a distance metric, D. The Euclidean distance is most frequently used as the metric.For a given network with M neurons and N-dimensional input vector x, the distance for neuron with weight vector Wj (j < M) is given by

Interoperable services based on activity monitoring in Ambient Assisted Living environments

“…The SOFM algorithm presented in [20] is based on a competitive learning algorithm, the winner-take-all (WTA) network, where an input vector is represented by the closest neuron prototype vector, which is assigned during training to a data cluster centre. The prototype vectors are stored in the "weights" of the neural network.…”

Binary object recognition system on FPGA with bSOM

“…Wei et al [15] presents an FPGA based a real-time face detection using AdaBoost algorithm. Appiah et al [5], presents a bSOM clustering algorithm and demonstrates it fast training rules on FPGAs. Lefebvre and Garcia [16] used SOM to measure image similarity in face recognition.…”

“…However, in many applications the data is either naturally presented as a binary string, or may be conveniently recoded as such (a "binary signature"). The bSOM [5] takes a binary vector input, and maintains tri-state vector weights with {0, 1, # } as the possible values. The # represents a "don't care" state, which signifies that the corresponding input vector bit may be either set or clear.…”

“…We use a tri-state rule binary Self Organising Map (bSOM) based on the system described in [5], modified to perform identification. The bSOM is capable of a wider variety of anomaly detection and classification tasks, justifying its use for identification among other tasks, while the FPGA based implementation makes it possible to design a single integrated on-chip system.…”

Metodologias para desenvolvimento de mapas auto-organizáveis de Kohonen executados em FPGA.

“…A área de pesquisa intitulada Engenharia de Sistemas Neurais (Neuroengineering) consiste no estudo, projeto e implementação de circuitos eletrônicos dedicados à execução de modelos neurais de computação TEMPLE, 2007 (APPIAH et al, 2009;APPIAH et al, 2010;APPIAH et al, 2012). Lachmair, que estuda o desenvolvimento de placas de circuito compostas por conjuntos de chips FPGAs para aceleração do processo de treinamento do SOM aplicado a operações de mineração de dados em grandes repositórios (LACHMAIR et al, 2012, LACHMAIR et al, 2013, LACHMAIR et al, 2017.…”