Computer scientists and immunologists collaborated on vertebrate immune system-inspired algorithms to create Artificial immune systems (AIS). It connects and fosters the disciplines of immunology, computer science, and engineering.

AIS applies immune system structure and function to computational architectures to solve mathematical, engineering, and information science analytical problems. This methodology is a subfield of biologically inspired and natural computation, including contributions to machine learning and the broader field of artificial intelligence. AIS is an adaptive system based on abstract immunology and resistive functions, with problem-solving principles and prototypes.

Overview

In the mid-1980s, Farmer et al. (1986) and Bersini et al. (1990) published publications on immunological networks that introduced AIS. Research is currently being done on new ideas related to AIS, such as hazard theory and algorithms based on the innate immune system. Although some argue that these new notions offer no truly "new" abstracts in addition to existing AIS methods, others disagree. It is, however, widely discussed, and the issue is currently one of the primary drivers pushing AIS growth. Another recent development is the examination of degeneracy in AIS models, inspired by its postulated involvement in open-ended learning and evolution.

Initially, AIS aimed to discover efficient abstractions of immune system operations, but more recently, it has become interested in simulating biological processes and applying immune algorithms to bioinformatics challenges. Furthermore, in 2008, Dasgupta and Nino published a handbook on immunological computation that compiles recent work on immunity-based techniques and explains a wide range of applications.

History

The origins of AIS may be traced back to the 1974 work of Jerne. His work demonstrates the philosophical component of the working of the immune network, which posits that immune system cells and molecules may recognize foreign substances, respond to foreign chemicals, and regulate one another. His hypothesis is known as immunological network theory. The researchers compared immune networks to brain networks. Ishida's paper describes the first attempt to employ the immunological network in problem-solving. It centred on creating distributed diagnosis systems based on immunological network interactions.

Inspired by how the immune system distinguishes between self (normal) and nonself (abnormal), an NSA was proposed. The work announced the path from immunology to computing. More AIS algorithms began to emerge due to the considerable advantages of immune-inspired algorithms while handling various issues in many application domains. The algorithms associated with the proposition are:

• the clonal selection algorithm (CLONAG), 
• artificial immune network algorithm (AINE), 
• danger theory-inspired algorithms, and 
• dendritic cell algorithm (DCA) 

Conclusion

The immune system is widely dispersed, adaptable, self-organizing, remembers previous interactions, and continuously learns from new ones. The AIS is an example of a system that uses current knowledge of the immune system. It demonstrates how AIS can imitate the immune system's fundamental components and exhibit some of its most prominent characteristics. Many aspects of natural immune systems, such as diversity, distributed computation, mistake tolerance, dynamic learning and adaptation, and self-monitoring, can be included in artificial immune systems. In addition, the human immune system has inspired scientists and engineers to develop practical information-processing algorithms, allowing them to address complicated engineering problems. 

AIS is a generic framework for a distributed adaptive system that, in theory, may be applied to various disciplines. For example, AIS applies to categorization issues, optimization tasks, and other fields. Like many systems inspired by biology, it is adaptable, distributed, and autonomous. The key benefits of AIS are that it requires only positive examples, and we may openly evaluate its learned patterns. Moreover, as it is self-organizing, no effort is needed to maximize system parameters.