Symbolic Learning Techniques Research Articles

A New Method for Earth Observation Data Analytics Based on Symbolic Machine Learning

This work introduces a new classification method in the remote sensing domain, suitably adapted to dealing with the challenges posed by the big data processing and analytics framework. The method is based on symbolic learning techniques, and it is designed to work in complex and information-abundant environments, where relationships among different data layers are assessed in model-free and computationally-effective modalities. The two main stages of the method are the data reduction-sequencing and the association analysis. The former refers to data representation; the latter searches for systematic relationships between data instances derived from images and spatial information encoded in supervisory signals. Subsequently, a new measure named the evidence-based normalized differential index, inspired by the probability-based family of objective interestingness measures, evaluates these associations. Additional information about the computational complexity of the classification algorithm and some critical remarks are briefly introduced. An application of land cover mapping where the input image features are morphological and radiometric descriptors demonstrates the capacity of the method; in this instructive application, a subset of eight classes from the Corine Land Cover is used as the reference source to guide the training phase.

Mining simulation data by rule induction to determine critical source areas of stream water pollution by herbicides

Modelling in environmental sciences is becoming increasingly complex because ever-increasing numbers of processes are combined, thus making model-based decision aids both more relevant but more difficult to develop. Our approach focused on water quality and aimed to identify spatial tree patterns that represent surface flow and pollutant pathways from plot to plot involved in water pollution by herbicides. First, a simulation model predicted herbicide transfer rate, the proportion of applied herbicide that reaches water courses, based on the spatial and temporal distribution of weed-control activities. These predictions were used as a set of learning examples for symbolic learning techniques to induce rules based on qualitative and quantitative attributes and explain two classes of transfer rate. In this study we compared two automatic symbolic learning techniques applied to a set of simulations: (1) a relational-inductive method using the inductive logic programming (ILP) approach to induce spatial tree patterns; and (2) an attribute-value method to induce aggregated attributes of the trees. Twenty-eight and thirty-three rules were learnt by the ILP and attribute-value approaches which explained 81% and 88% of the examples, respectively. The ILP approach provided relevant indicators of plot-to-plot connectivity. The integrated attribute-value approach is simpler and quicker, but the ILP patterns are more useful for stakeholders.

Scalability in Formal Concept Analysis

Formal Concept Analysis is a symbolic learning technique derived from mathematical algebra and order theory. The technique has been applied to a broad range of knowledge representation and exploration tasks in a number of domains. Most recorded applications of Formal Concept Analysis deal with a small number of objects and attributes, in which case the complexity of the algorithms used for indexing and retrieving data is not a significant issue. However, when Formal Concept Analysis is applied to exploration of a large numbers of objects and attributes, the size of the data makes issues of complexity and scalability crucial.This paper presents the results of experiments carried out with a set of 4,000 medical discharge summaries in which were recognized 1,962 attributes from the Unified Medical Language System (UMLS). In this domain, the objects are medical documents (4,000) and the attributes are UMLS terms extracted from the documents (1,962). When Formal Concept Analysis is used to iteratively analyze and visualize these data, complexity and scalability become critically important.Although the amount of data used in this experiment is small compared with the size of primary memory in modern computers, the results are still important because the probability distributions that determine the efficiencies are likely to remain stable as the size of the data is increased.Our work presents two outcomes. First, we present a methodology for exploring knowledge in text documents using Formal Concept Analysis by employing conceptual scales created as the result of direct manipulation of a line diagram. The conceptual scales lead to small derived purified contexts that are represented using nested line diagrams. Second, we present an algorithm for the fast determination of purified contexts from compressed representation of the large formal context. Our work draws on existing encoding and compression techniques to show how rudimentary data analysis can lead to substantial efficiency improvements in knowledge visualization.

Machine learning of material behaviour knowledge from empirical data

Symbolic machine learning techniques can extract flexible and comprehensible knowledge from empirical data of material behaviour. The diversity of symbolic machine learning techniques offers potential to match the requirements of many tasks when models of material behaviour need to be created from data. We develop a series of steps for generating material behaviour knowledge from empirical data and exemplify some of them on several small datasets. We discuss some of the issues that govern knowledge extraction and, as a by-product, demonstrate that symbolic learning techniques are functionally superior to sub-symbolic learning for the task of comprehensible knowledge extraction.