Liquid drop photonic signal analysis using fast learning artificial neural networks

This paper presents a treatment on data obtained from a liquid drop photonic signal analyzer. The liquid drop analyzer extracts liquid features from different types of liquid drops and obtains a spectrum of characteristics. The data is then clustered using the K-means fast learning artificial neural network (K-FLANN) that implements a systematic reshuffling of the input data points to achieve consistent clustering, regardless of the data input sequence. An introduction of the K-FLANN network is presented in this paper as it is rarely encountered. The discussions explains how the K-FLANN stabilizes the cluster formations such that the resultant cluster centroids remain relatively consistent even though the clustering is done on data presented in a different sequence. The experimental results have a dual agenda. Firstly it shows that the liquid drop photonic data is a viable method of discriminating between liquids and secondly the K-FLANN is resilient changes in data presentation sequences and preserves the clustering consistencies.