No description available.

No description available.

No description available.

This thesis addresses the question: "How can knowledge learnt through inductive inference be communicated in a multi-agent system?". Existing agent communication languages, such as KQML, assume logically sound inference methods. Unfortunately, induction is logically unsound. In general, machine learning techniques infer knowledge (or hypotheses) consistent with the locally available facts. However, in a multi-agent system, hypotheses learnt by one agent can directly contradict knowledge held by another. If an agent communicates induced knowledge as though it were logically sound, then the knowledge held by other agents in the community may become inconsistent. The answer we present in this thesis is that agents must, in general, communicate the bounds to such induced knowledge. The Version Space framework characterises inductive inference as a process which identifies the set of hypotheses that are consistent with both the observable facts and the constraints of the hypothesis description language. A Version Space can be expressed by two boundary sets, which represent the most general and most specific hypotheses. We thus propose that when communicating an induced hypothesis, that the hypothesis be bounded by descriptions of the most general and most specific hypotheses. In order to allow agents to integrate induced hypotheses with their own facts or their own induced hypotheses, the technique of Version Space Intersection can be used. We have investigated how boundary set descriptions can be generated for the common case of machine learning algorithms which learn hypotheses from unrestricted Version Spaces. This is a hard computational problem, as it is the equivalent of finding the minimal DNF description of a set of logical sentences. We consider four alternate approaches: exact minimization using the Quine-McCluskey algorithm; a naive, information-theoretic hill-climbing search; Espresso II, a sophisticated, heuristic logic minimization algorithm; and unsound approximation techniques. We demonstrate that none of these techniques are scalable to realistic machine learning problems.

No description available.

No description available.

No description available.

No description available.

No description available.

This thesis develops computational memory architectures for autobiographic and narrative virtual agents. Humans and many animals naturally possess a sophisticated memory system for reasoning, learning and also sharing information with others. However it has been a difficult challenge to model the characteristics of such a memory system in the research fields of both Artificial Intelligence and Artificial Life. We propose a framework for enhancing reactive autonomous agents to retrieve meaningful information from their dynamic memories in order to adapt and survive in their environments. Our approach is inspired by psychology research in human memory and autobiographic memory – through remembering the significance of episodic events that happened in the past, agents with autobiographic memory architectures are capable of reconstructing past events for the purpose of event re-execution and story-telling. The memory architectures that were developed are capable of organizing and filtering significant events which originate in agents’ own experiences as well as stories told by other agents. To validate our memory architectures, both simple and complex Artificial Life type of virtual environments with static as well as dynamic resources distribution were implemented that provide events with different levels of complexity and affect the internal variables of the agents. The performance of various types of agents with different memory control architectures are first compared in single-agent experiments. Each agent’s behaviour is observed and analysed quantitatively together with its lifespan and internal states measurements. Group performance with and without communication are measured in experiments with multiple autobiographic ii agents. Results confirm our research hypothesis that autobiographic memory can prove beneficial – resulting in increases in the lifespan of an autonomous, autobiographic, minimal agent. Furthermore, higher communication frequency brings better group performance for Long-term Autobiographic Memory agents in multi-agent experiments. An interface has been developed to visualise agents’ dynamic autobiographic memory to help human observers to understand the underlying memory processes. This research leads to insights into how bottom-up story-telling and autobiography reconstruction in artificial autonomous agents allow temporally grounded behaviour to emerge. This study therefore results in a contribution to knowledge in Artificial Life and Artificial Intelligence.