Search

Recent years have seen a renewed interest in evolutionary computation applied to the automatic design of deep neural network architectures, i.e. Neural Architecture Search (NAS). The advantages of evolutionary approaches in NAS include their conceptual simplicity and their flexibility with regards to search space definition and/or optimization objective.However, Artificial Immune Systems (AIS) that follow the evolutionary computation paradigm are less explored in NAS. In this research, we aim to leverage their intrinsic and excellent ability to balance performance and population diversity to develop a novel Neural Ensemble Search method, based on the Clonal Selection Algorithm [1]. For more generality, we focus on designing macro-architectures rather than architectural components.Experiments on popular computer vision benchmarks demonstrate that our method reaches competitive accuracy and efficiency despite minimal augmentation and post-processing. We show that the AIS brings tangible benefits, including maintaining the diversity of solutions, a semantically straightforward implementation, and high efficiency. Moreover, this AIS can exhibit a "secondary response": when presented with a related but more difficult task, the ensemble will perform competently with zero modification to the architectures or the training protocol.

Missing data is challenging enough without the added complexities posed by a lack of research in evaluating imputation. Not only could we potentially increase the impact and validity of studies from many different sectors (research, public and private), we also believe that by creating evaluation software, more researchers may be willing to use and justify using imputation methods. This paper aims to encourage further research for efficient imputation evaluation by defining a framework which could be used to optimise the way we impute datasets prior to data analysis. We propose a framework which uses a prototypical approach to create testing data and machine learning methods to create a new metric for evaluation. Preliminary results are presented which show how, for our dataset, records with less than 40% missingness could be used for analysis, increasing the amount of available data.

In this paper, a novel qualitative differential equation model learning (QML) framework named QML-Morven is presented. QML-Morven employs both symbolic and evolutionary approaches as its learning strategies to deal with models of different complexity. Based on this framework, a series of experiments were designed and carried out to: (1) investigate factors that influence the learning precision and minimum data requirement for successful learning; (2) address the scalability issue of QML systems.

The Artificial Reaction Network (ARN) is a Cell Signalling Network inspired connectionist representation belonging to the branch of A-Life known as Artificial Chemistry. Its purpose is to represent chemical circuitry and to explore computational properties responsible for generating emergent high-level behaviour associated with cells. In this paper, the computational mechanisms involved in pattern recognition and spatio-temporal pattern generation are examined in robotic control tasks. The results show that the ARN has application in limbed robotic control and computational functionality in common with Artificial Neural Networks. Like spiking neural models, the ARN can combine pattern recognition and complex temporal control functionality in a single network, however it offers increased flexibility. Furthermore, the results illustrate parallels between emergent neural and cell intelligence.

A reliable human skin detection method that is adaptable to different human skin colors and illumination conditions is essential for better human skin segmentation. Even though different human skin-color detection solutions have been successfully applied, they are prone to false skin detection and are not able to cope with the variety of human skin colors across different ethnic. Moreover, existing methods require high computational cost. In this paper, we propose a novel human skin detection approach that combines a smoothed 2-D histogram and Gaussian model, for automatic human skin detection in color image(s). In our approach, an eye detector is used to refine the skin model for a specific person. The proposed approach reduces computational costs as no training is required, and it improves the accuracy of skin detection despite wide variation in ethnicity and illumination. To the best of our knowledge, this is the first method to employ fusion strategy for this purpose. Qualitative and quantitative results on three standard public datasets and a comparison with state-of-the-art methods have shown the effectiveness and robustness of the proposed approach.

In this paper, a special-purpose qualitative model learning (QML) system using an immune-inspired algorithm is proposed to qualitatively reconstruct biological pathways. We choose a real-world application, the detoxification pathway of Methylglyoxal (MG), as a case study. First a converter is implemented to convert possible pathways to qualitative models. Then a general learning strategy is presented. To improve the scalability of the proposed QML system and make it adapt to future more complicated pathways, a modified clonal selection algorithm (CLONALG) is employed as the search strategy. The performance of this immune-inspired approach is compared with those of exhaustive search and two backtracking algorithms. The experimental results indicate that this immune-inspired approach can significantly improve the search efficiency when dealing with some complicated pathways with large-scale search spaces.

Experience in the physical sciences suggests that the only realistic means of understanding complex systems is through the use of mathematical models. Typically, this has come to mean the identification of quantitative models expressed as differential equations. Quantitative modelling works best when the structure of the model (i.e., the form of the equations) is known; and the primary concern is one of estimating the values of the parameters in the model. For complex biological systems, the model-structure is rarely known and the modeler has to deal with both model-identification and parameter-estimation. In this paper we are concerned with providing automated assistance to the first of these problems. Specifically, we examine the identification by machine of the structural relationships between experimentally observed variables. These relationship will be expressed in the form of qualitative abstractions of a quantitative model. Such qualitative models may not only provide clues to the precise quantitative model, but also assist in understanding the essence of that model. Our position in this paper is that background knowledge incorporating system modelling principles can be used to constrain effectively the set of good qualitative models. Utilising the model-identification framework provided by Inductive Logic Programming (ILP) we present empirical support for this position using a series of increasingly complex artificial datasets. The results are obtained with qualitative and quantitative data subject to varying amounts of noise and different degrees of sparsity. The results also point to the presence of a set of qualitative states, which we term kernel subsets, that may be necessary for a qualitative model-learner to learn correct models. We demonstrate scalability of the method to biological system modelling by identification of the glycolysis metabolic pathway from data. ©2008 AI Access Foundation. All rights reserved.

Modelling of biochemical pathways in a computational way has received considerable attention over the last decade from biochemistry, computing sciences, and mathematics. In this paper we present an approach to evolutionarily stepwise constructing models of biochemical pathways by a qualitative model learning methodology. Given a set of reactants involved in a target biochemical pathway, atomic components can be generated and preserved in a components library for further model composition. These synthetic components are then reused to compose models which are qualitatively evaluated by referring to experimental qualitative states of the given reactants. Simulation results show that our stepwise evolutionary qualitative model learning approach can learn the relationships among reactants in biochemical pathway, by exploring topology space of alternative models. In addition, synthetic biochemical complex can be obtained as hidden reactants in composed models. The inferred hidden reactants and topologies of the synthetic models can be further investigated by biologists in experimental environment for understanding biological principles.

The Artificial Reaction Network (ARN) is a bio-inspired connectionist paradigm based on the emerging field of Cellular Intelligence. It has properties in common with both AI and Systems Biology techniques including Artificial Neural Networks, Petri Nets, and S-Systems. This paper discusses the temporal aspects of the ARN model using robotic gaits as an example and compares it with properties of Artificial Neural Networks. The comparison shows that the ARN based network has similar functionality.

In this paper, an Evolutionary Qualitative Model LearningFramework (EQML) is proposed and tested by learning the qualitative metabolic models under the condition of incomplete knowledge. JMorven, a fuzzy qualitative reasoning engine, is slightly modified and integrated into the framework as a sub-module to represent and verify the candidate models. Three metabolic compartment models are tested by two evolutionary algorithms (Genetic Algorithm and Clonal Selection Algorithm) in EQML. Finally the efficiency of these two algorithms is evaluated.

In this paper we continue the research on applying immune-inspired algorithms as search strategies to Qualitative Model Learning (QML). A new search strategy based on opt-AiNet is proposed, and this results in the development of a novel QML system called QML-AiNet. The performance of QML-AiNet is compared with previous work using the CLONALG approach. Experimental results shows that although not as efficient as CLONALG, the opt-AiNet based approach still shows promising results for learning qualitative models. In addition, possible future work to further improve the efficiency of QML-AiNet is also pointed out.

In this paper, a qualitative model learning (QML) system is proposed to qualitatively reconstruct the detoxification pathway of Methylglyoxal. First a converting algorithm is implemented to convert possible pathways to qualitative models. Then a general learning strategy is presented. To improve the scalability of the proposed QML system and make it adapt to future more complicated pathways, an immune-inspired approach, a modified clonal selection algorithm, is proposed. The performance of this immune-inspired approach is compared with those of exhaustive search and two backtracking algorithms. The experimental results indicate that this approach can significantly improve the search efficiency when dealing with some complicated pathways with large-scale search spaces.

We present a qualitative model-learning system, Qoph , developed for application to scientific discovery problems. Qoph learns the structuralrelations between a set of observed variables. It has been shown capable of learning models with intermediate (unmeasured) variables, and intermediate relations, under different levels of noise, and from qualitative or quantitative data. A biological application of Qoph is explored. An additional significant outcome of this work is the discovery and identification of kernel subsets of key states that must be present for model-learning to succeed.

In this paper, a modified Clonal Selection Algorithm (CSA)is proposed to learn qualitative compartmental models. Different from traditional AI search algorithm, this population based approach employs antibody repertoire to perform random search, which is suitable for the ragged and multi-modal landscape of qualitative model space. Experimental result shows that this algorithm can obtain the same kernel sets and learning reliability as previous work for learning the two compartment model, and it can also search out the target model when learning the more complex three-compartment model. Although this algorithm does not succeed in learning the four-compartment model, promising result is still obtained.

This paper is concerned with the learning of dynamic models of compartmental systems visualized as networks of interconnected tanks. This is intended as an intermediary step to learn more complex dynamic biological systems such as metabolic pathways. Our present aim is to learn systems of differential equations from time series data to capture physical models of increasing complexity (u-tube, cascaded tanks, and coupled tanks). To do so, we use Symbolic Regression in Genetic Programming and combine it with a fuzzy representation which has inherent differential capabilities (Fuzzy Vector Envisionment). We use the ECJ1 framework to implement the learner. Present results show that the system can approximate the target models and that the use of a weighted fitness function seems to accelerate the learning process.

Fuzzy qualitative simulation combines the features of qualitative simulation and fuzzy reasoning in order to gain advantages from both. However, the output of a fuzzy qualitative simulation process is a behaviour tree which for complex systems will be large. In order to overcome this and permit focussing on preferred behaviours priortisation was developed. In this paper a new prioritisation scheme is presented that makes use of both constraint and temporal information to perform the prioritisation.