Artificial Immune Systems

Leishmania donovani

is an obligate intracellular parasite responsible for the systemic disease visceral leishmaniasis. During the course of the disease, the parasite is found in the spleen, liver and bone marrow. Characteristic of the liver immune response to leishmaniasis is a type of inflammation (“ggranulomatous inflammation”) that results in the formation of granulomas, structures comprised of an infiltrate of mononuclear cells surrounding a core of infected macrophages. Granulomas help limit the spread of infection and facilitate the killing of parasites.

Liver-resident macrophages (Kupffer cells) are able to spontaneously kill many infectious agents, but

L. donovani

is capable of reproducing inside these cells. Activation of Kupffer cells is required to turn them from host cell to a cell that is able to kill intracellular

L. donovani

. This process of activation is regulated by cytokines (notably IFN

γ

) produced by many different types of leukocytes, including natural killer (NK) cells ([1]), CD4

 + 

and CD8

 + 

T cells ([2]), and NKT cells ([3]).

Although there are various types of cancer treatments, none of these currently take into account the effect of ageing of the immune system and hence altered responses to cancer. Recent studies have shown that

in vitro

stimulation of T cells can help in the treatment of patients. There are many factors that have to be considered when simulating an organism’s immunocompetence. Our particular interest lies in the study of loss of immunocompetence with age. We are trying to answer questions such as: Given a certain age of a patient, how fit is their immune system to fight cancer? Would an immune boost improve the effectiveness of a cancer treatment given the patient’s immune phenotype and age? We believe that understanding the processes of immune system ageing and degradation through computer simulation may help in answering these questions. Specifically, we have decided to look at the change in numbers of naive T cells with age, as they play a important role in responses to cancer and anti-tumour vaccination. In this work we present an agent-based simulation model to understand the interactions which influence the naive T cell populations over time. Our agent model is based on existing mathematical system dynamic model, but in comparisons offers better scope for customisation and detailed analysis. We believe that the results obtained can in future help with the modelling of T cell populations inside tumours.

Clonal selection is the keystone of mainstream immunology and computational systems based on immunological principles. For the latter, clonal selection is often interpreted as an asexual variant of natural selection, and thus, tend to be variations on evolutionary strategies. Retro-fitting immunological sophistication and theoretical rigour onto such systems has proved to be unwieldy. In this paper we assert the primacy of

competitive exclusion

over selection and mutation; providing theoretical analysis and empirical results that support our position.

We show our algorithm to be highly competitive with well-established approximation and learning techniques; notably for large, high-dimensional problems. We further argue that it provides concise, operational abstractions of some influential theoretical immunology.

A clustering algorithm may be designed to generate prototypes capable of minimizing the cumulative distance between each sample in the dataset and its corresponding prototype, denoted as minimum quantization error clustering. On the other hand, some clustering applications may require density-preserving prototypes, more specifically prototypes that maximally obey the original density distribution of the dataset. This paper presents a conceptual framework to demonstrate that both criteria are attainable but are distinct and cannot be fulfilled simultaneously. Illustrative examples are used to validate the framework, further applied to produce an adaptive radius immune-inspired algorithm capable of transiting between both criteria in practical applications.

Adaptive Information Filtering is a challenging computational problem that requires a high dimensional feature space. However, theoretical issues arise when vector-based representations are adopted in such a space. In this paper, we use AIF as a test bed to provide experimental evidence indicating that the learning abilities of vector-based Artificial Immune Systems are diminished in a high dimensional space.

Aging is a concept that is used in many artificial immune system implementations. It is an important tool that helps to cope with multi-modal problems by increasing diversity and allowing to restart the search in different parts of the search space. The current theoretical understanding of the details of aging is still very limited. This holds with respect to parameter settings, the relationship of different variants, the specific mechanisms that make aging useful, and implementation details. While implementation details seem to be the least important part they can have a surprisingly huge impact. This is proven by means of theoretical analysis for a carefully constructed example problem as well as thorough experimental investigations of aging for this problem.

The dendritic cell algorithm is often presented as an immune-inspired one class classifier. Recently the dendritic cell algorithm has been criticised as its current decision making stage has many serious mathematical flaws which bring into question its applicability in other areas. However, previous work has demonstrated that the algorithm has properties which make it robust to a certain source of uncertainty, specifically measurement noise. This paper presents a discussion about the role of uncertainty within classification tasks and goes on to identify the strengths and weaknesses of the dendritic cell algorithm from this perspective. By examining other techniques for protecting against uncertainty, future directions for the dendritic cell algorithm are identified and discussed.

The aim of this paper is to investigate the antigen sampling component of the deterministic version of the dendritic cell algorithm (dDCA). To achieve this, a model is presented, and used to produce synthetic data for two temporal correlation problems. The model itself is designed to simulate a system stochastically switching between a normal and an anomalous state over time. By investigating five parameter values for the dDCA’s maximum migration threshold, and benchmarking alongside a minimised version of the dDCA, the effect of sampling using a multi-agent population is explored. Potential sources of error in the dDCA outputs are identified, and related to the duration of the anomalous state in the input data.

An immune-inspired danger theory model based on dendritic cells (DCs) within the framework of fuzzy set theory is proposed in this paper. Our objective is to smooth the abrupt separation between normality (semi-mature) and abnormality (mature) using fuzzy set theory since we can neither identify a clear boundary between the two contexts nor quantify exactly what is meant by “semi-mature” or “mature”. In this model, the context of each object (DC) is described using linguistic variables. Fuzzy subsets and the corresponding membership functions describe these variables. A knowledge base, comprising rules, is built to support the fuzzy inference. The induction of the context of each object is diagnosed using a compositional rule of fuzzy inference. Experiments on real data sets show that by alleviating the crisp separation between the two contexts, our new approach which focuses on binary classification problems produces more accurate results.

The Natural Immune System (NIS) is a distributed system that solves challenging search and response problems while operating under constraints imposed by physical space and resource availability. Remarkably, NIS search and response times do not scale appreciably with the physical size of the animal in which its search is conducted. Many distributed systems are engineered to solve analogous problems, and the NIS demonstrates how such engineered systems can achieve desirable scalability. We hypothesize that the architecture of the NIS, composed of a hierarchical decentralized detection network of lymph nodes (LN) facilitates efficient search and response. A sub-modular architecture in which LN numbers and size both scale with organism size is shown to efficiently balance tradeoffs between local antigen detection and global antibody production, leading to nearly

scale-invariant detection and response

. We characterize the tradeoffs as balancing local and global communication and show that similar tradeoffs exist in distributed systems like LN inspired artificial immune system (AIS) applications and peer-to-peer (P2P) systems. Taking inspiration from the architecture of the NIS, we propose a modular RADAR (Robust Adaptive Decentralized search with Automated Response) strategy for distributed systems. We demonstrate how two existing distributed systems (a LN inspired multi-robot control application and a P2P system) can be improved by a modular RADAR strategy. Such a sub-modular architecture is shown to balance the tradeoffs between local communication (within artificial LNs and P2P clusters) and global communication (between artificial LNs and P2P clusters), leading to efficient search and response.

Artificial Immune Systems (AISs) are computational methods, inspired by the biological immune system, that can be applied to solve optimization problems. In this paper we propose the use of a similarity-based surrogate model in conjunction with a clonal selection algorithm in order to improve its performance when solving optimization problems involving computationally expensive objective functions. Computational experiments to assess the performance of the proposed procedure using 23 test-problems from the literature are presented.

In this research work a large set of the classical numerical functions were taken into account in order to understand both the search capability and the ability to escape from a local optimal of a clonal selection algorithm, called

i-CSA

. The algorithm was extensively compared against several variants of

Differential Evolution

(DE) algorithm, and with some typical swarm intelligence algorithms. The obtained results show as

i-CSA

is effective in terms of accuracy, and it is able to solve large-scale instances of well-known benchmarks. Experimental results also indicate that the algorithm is comparable, and often outperforms, the compared nature-inspired approaches. From the experimental results, it is possible to note that a longer maturation of a B cell, inside the population, assures the achievement of better solutions; the maturation period affects the diversity and the effectiveness of the immune search process on a specific problem instance. To assess the learning capability during the evolution of the algorithm three different relative entropies were used:

Kullback-Leibler

,

Rényi generalized

and

Von Neumann

divergences. The adopted entropic divergences show a strong correlation between optima discovering, and high relative entropy values.

An adaptive radius immune algorithm proposed in the literature, denoted as ARIA, is claimed to preserve the density distribution of the original dataset when generating prototypes. Density-preserving prototypes may correspond to high-quality compact representations for clustering applications. The original samples in the dataset are interpreted as antigens, and the prototypes are interpreted as antibodies. In this paper, some theoretical results are provided to demonstrate that the original version of ARIA is not capable of generating density-preserving prototypes when high-dimensional datasets are considered. Further, the same theoretical results are explored to conceive a new version of ARIA, now capable of exhibiting the announced density-preserving attribute. The main innovation is in the way the algorithm estimates local densities.

This paper proposes a Gaussian Artificial Immune System (GAIS) to deal effectively with building blocks (high-quality partial solutions coded in the solution vector) in continuous optimization problems. By replacing the mutation and cloning operators with a probabilistic model, more specifically a Gaussian network representing the joint distribution of promising solutions, GAIS takes into account the relationships among the variables of the problem, avoiding the disruption of already obtained high-quality partial solutions. Two versions of the algorithm were developed. In the first one, the estimation of the joint probability distribution is achieved by means of a single multivariate Gaussian distribution. In the second version, the estimation is carried out using a Gaussian mixture model. The algorithms were applied to eight benchmarks and the results compared with those produced by an immune-inspired algorithm and an estimation of distribution algorithm.

Wireless mesh networks typically employ mesh routers that are equipped with multiple radio interfaces to improve network capacity. The key is to cleverly assign different channels (i.e. frequency bands) to each radio interface to form a mesh network with minimum interference. The channel assignment must obey the constraints that the number of different channels assigned to a router is at most the number of interfaces on the router, and the resultant mesh network is connected. This problem is known to be NP-hard. We propose an immune algorithm which performs greedy channel assignment to obtain feasible solutions, and use the clonal selection principle to improve the solutions. Through extensive simulations, we show that our algorithm out-perform a genetic algorithm and a graph-theoretic algorithm proposed for the same problem.

This paper presents a high level dynamic task allocation algorithm that is inspired by the biological development process and the immune system. For a microprocessor (

μ

P) array, a program is partitioned into a number of workload oriented tasks with data dependencies and a number of internal status-oriented tasks. Each

μ

P in the array is capable of processing one of these tasks. The algorithm assigns tasks to the

μ

P array that satisfies the requirements of the problem, and it dynamically recovers the system from faults at runtime.

The doping profile optimization in semiconductor has been tackled as a constrained optimization problem coupled with a drift-diffusion model to simulate the physical phenomenon. In order to design high performance semiconductor devices, a new immunological algorithm, the Constrained Immunological Algorithm (cIA), has been introduced. The experimental results confirm that

cIA

clearly outperforms previous state-of-the-art algorithms in doping profile optimization.

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.

We propose a novel bio-inspired solution for biomedical article classification. Our method draws from an existing model of T-cell cross-regulation in the vertebrate immune system (IS), which is a complex adaptive system of millions of cells interacting to distinguish between harmless and harmful intruders. Analogously, automatic biomedical article classification assumes that the interaction and co-occurrence of thousands of words in text can be used to identify conceptually-related classes of articles—at a minimum, two classes with relevant and irrelevant articles for a given concept (e.g. articles with protein-protein interaction information). Our agent-based method for document classification expands the existing analytical model of Carneiro et al. [1], by allowing us to deal simultaneously with many distinct T-cell features (epitomes) and their collective dynamics using agent based modeling. We already extended this model to develop a bio-inspired spam-detection system [2, 3]. Here we develop our agent-base model further, and test it on a dataset of publicly available full-text biomedical articles provided by the BioCreative challenge [4]. We study several new parameter configurations leading to encouraging results comparable to state-of-the-art classifiers. These results help us understand both T-cell cross-regulation and its applicability to document classification in general. Therefore, we show that our bio-inspired algorithm is a promising novel method for biomedical article classification and for binary document classification in general.

An artificial immune system approach for artificial chemistries (ACs) based on set rewriting methods is introduced. We model signals which are generated by the execution of rewriting rules in the artificial chemistry. They induce signal patterns which trigger a system response. This response is also based on the rewriting rules of the AC. The latter inhibit or accelerate self-reproducing rewriting instructions and eliminate or inhibit non-self rewriting rules in the AC. Exemplarily, the developed artificial immune system model is integrated into the computational model of Fraglets, an AC which is based on multiset rewriting. Experimental results show the feasibility of this approach.

This paper presents continued experimentation on the Network Threat Recognition with Immune Inspired Anomaly Detection, or NetTRIIAD, model. This hybrid model combines established network monitoring methods with artificial immune system methods to achieve improved performance. The paper presets experiments investigating the model’s performance in detecting novel threats and the performance contribution of the individual components.

Metaphors derived from Danger Theory, a hypothesized model of how the human immune system works, have been applied to the intrusion detection domain. The major contribution in this area, is the dendritic cell algorithm (DCA). This paper presents an in-depth analysis of results obtained from two previous experiments, regarding the suitability of the danger theory analogy in constructing intrusion detection systems for web applications. These detectors would be capable of detecting novel attacks while improving on the limitations of anomaly-based intrusion detectors. In particular, this analysis investigates which aspects of this analogy are suitable for this purpose, and which aspects of the analogy are counterproductive if utilized in the way originally suggested by danger theory. Several suggestions are given for those aspects of danger theory that are identified to require modification, indicating the possibility of further pursuing this approach. These modifications could be realized in terms of developing a robust signal selection schema and a suitable correlation algorithm. This would allow for an intrusion detection approach that has the potential to overcome those limitations presently associated with existing techniques.

This paper proposes an improved version of current electronic fraud detection system by using logging data sets for Video-on-Demand system. Our approach is focused on applying Artificial Immune System based fraud detection algorithm for logging data information and accounting and billing purposes. Our hybrid approach combines algorithms from innate and adaptive parts of immune system, inspired by the Self non-self theory and the Danger theory. Our research proved the possibility of combining these to perform E-fraud detection. The experimental results demonstrated that hybrid approach has higher detection rate, lower false alarm when compared with the performances achieved by traditional classification algorithms such as Decision Tree, Support Vector Machines, and Radial Basis Function Neural Networks. Our approach also outperforms AIS approaches that use Dendritic Cell Algorithm, Conserved Self Pattern Recognition Algorithm, and Clonal Selection Algorithm individually.

This position statement presents a brief overview of our research programme investigating the convergence of biologically-inspired robotics with sociologically inspired agents, and its potential application in multi-dimensional intelligent transportations systems.

The biological term Homeostasis refers to an organisms ability to maintain steady states of operation in a massively changing internal and external environment. Pervasive Adaptive (PerAda) systems operate in potentially hostile environments where there is a need to cope with a wide variety of changes, over multiple timescales, deal with interacting agents, cope with degraded data and cope with failing components. In the context of this paper, we consider the case of a swarm, and organism, of robotic units, as those developed as part of the SYMBRION project. Such a collective robotic system can be considered a case of a PerAda system.

Our position is that for such systems to operate for extended periods of time, without human intervention, they should be

homeostatic

and that it is possible to develop such homeostatic PerAda systems inspired by the immune system

.

The major challenge to multi-cellular robotics system is how to ensure the system is homeostatically stable. This position paper proposes a developmental artificial immune system (dev-AIS) framework that tries to provide and maintain homeostasis to the multi-cellular robotics system. If

immunity

is defined as the ability to maintain homeostasis; the dev-AIS framework will be designed based on the understanding and the abstraction of how different organisms attain for this property through evolution and developmental process. Progress in evolution drove the evolution of

immunity

from this simple relationship to the development of the immune system in multi-cellular organisms.

One possibility to increase the efficiency of classical controller paradigms is to implement substructures with different tasks and abilities (heterogeneous controller structure). To automatically develop these structures it is necessary to use a process, that is on the one hand easy to evolve, and on the other hand rich in different solutions. One inspiration for such processes can be found in the biological process of embryogenesis. By using virtual embryogenetical processes in combination with artificial evolution we develop a novel method for controller structuring.

Pervasive Adaptation (PerAda) refers to massive-scale pervasive information and communication systems which are capable of autonomously adapting to highly dynamic and open technological and user contexts. PerAda systems are thus a special case of collective adaptive systems which have particular constraints e.g. they are networked and highly distributed; they involve interaction with humans; they are large scale; the boundaries of systems are fluid; their context is dynamic; and they operate using uncertain information. In order to achieve their ultimate goal of adapting seamlessly to their users and to deliver the expected quality of service at all times, we propose that these systems must exhibit self-awareness. This position statement proposes mechanisms by which self-awareness might be achieved.

Wireless ad hoc networks are an example for pervasive communication systems. Due to their decentralized design, they can be subject to a large variety of communication errors, system failures or even attacks and intrusions. Current research focuses either on specific avoidance and/or protective actions such as secure routing protocols or on general misbehavior detection with human operator based response management [1]. However, due to the complexity of large-scale ad hoc networks, self-organization and self-healing properties are necessary. A concept which allows for automatic responses to mitigate the effects of misbehavior is missing. In this statement we discuss whether an adaptive decision making process based on the concept of

receptor degeneracy

, being a part of the

cognitive immune self

theory described by Cohen in 2000, is feasible.

Adaptation software systems are expected to exhibit life-like properties such as robust operation in uncertain environments, adaptive immunity against foreign attackers, self-maintenance, and so on. The traditional software design model based on top-down human engineering fails in this context, where new, bottom-up

emergent computation

[1,2] techniques seem more appropriate.

Since chemistry and biochemistry are the basis of life,

Artificial Chemistries

[3] and

Artificial Biochemistries

[4] stand out as natural ways to model such bottomup life-like software. However, understanding and harnessing the power of emergent behavior in such complex systems is difficult. This position statement highlights some potentially fruitful research directions towards this goal. We advocate that an important research goal within such bottom-up approach is to construct systems able to achieve automatic transitions from lower levels of complexity to higher ones.

Paradigms like delay-tolerant networks or opportunistic networking appeared as networked environments called for new approaches in communication. Due to the often decentralized and self-organized character of these systems, the amount of explicit control in these systems must be reduced radically.

Our vision moves one step further, towards a system where the

communication scheme itself

also emerges from the actual conditions. When each element of the network uses a simple mechanism to fine-tune the amount and way of communication to match its local needs and resources, we can easily end up in an emergent system where the communication (amount and content) is globally adaptive, ’by itself’.

We describe the concept of a cognitive network and propose that ecosystems of co-existing networks which are globally energy efficient while providing the expected quality of service can be realised by exploiting two mechanisms which occur in biological systems; symbiosis and co-existence.