Neural architecture search (NAS) has demonstrated impressive performance in automatically designing high-performance neural networks. The power of deep neural networks is to be unleashed for analyzing a large volume of data (e.g. ImageNet), but the architecture search is often executed on another smaller dataset (e.g. CIFAR-10) to finish it in a feasible time. However, it is hard to guarantee that the optimal architecture derived on the proxy task could maintain its advantages on another more challenging dataset. This paper aims to improve the generalization of neural architectures via domain adaptation. We analyze the generalization bounds of the derived architecture and suggest its close relations with the validation error and the data distribution distance on both domains. These theoretical analyses lead to AdaptNAS, a novel and principled approach to adapt neural architectures between domains in NAS. Our experimental evaluation shows that only a small part of ImageNet will be sufficient for AdaptNAS to extend its architecture success to the entire ImageNet and outperform state-of-the-art comparison algorithms.

Various real-world challenges require planning algorithms that can adapt to a broad range of domains. Traditionally, the creation of planning domains has relied heavily on human implementation, which limits the scale and diversity of available domains. While recent advancements have leveraged generative AI technologies such as large language models (LLMs) for domain creation, these efforts have predominantly focused on translating existing domains from natural language descriptions rather than generating novel ones. In contrast, the concept of domain randomization, which has been highly effective in reinforcement learning, enhances performance and generalizability by training on a diverse array of randomized new domains. Inspired by this success, our tool, PDDLFuse, aims to bridge this gap in Planning Domain Definition Language (PDDL). PDDLFuse is designed to generate new, diverse planning domains that can be used to validate new planners or test foundational planning models. We have developed methods to adjust the domain generators parameters to modulate the difficulty of the domains it generates. This adaptability is crucial as existing domain-independent planners often struggle with more complex problems. Initial tests indicate that PDDLFuse efficiently creates intricate and varied domains, representing a significant advancement over traditional domain generation methods and making a contribution towards planning research.

Is Robot Learning a New Subfield? The third section argues that the machine-learning and robotics communities reflect different cultures, target domains, terminology, and acceptable proofs that result in a de facto separation. The unique constraints placed on representation by robot learning are characterized in the fourth section. Finally, we close with some concluding remarks. Learning takes place when the system makes changes to its internal structure so as to improve some metric on its long-term future performance, as measured by a fixed standard (Russell 1991, p. 141).

We use our experience with the Dipmeter Advisor system for well-log interpretation as a case study to examine the development of commercial expert systems. We discuss the nature of these systems as we see them in the coming decade, characteristics of the evolution process, development methods, and skills required in the development team. We argue that the tools and ideas of rapid prototyping and successive refinement accelerate the development process. We note that different types of people are required at different stages of expert system development: Those who are primarily knowledgeable in the domain, but who can use the framework to expand the domain knowledge; and those who can actually design and build expert system tools and components We also note that traditional programming skills continue to be required in the development of commercial expert systems Finally, we discuss the problem of technology transfer and compare our experience with some of the traditional wisdom of expert system development. We have observed during this effort that the development of a commercial expert system imposes a substantially different set of constraints and requirements in terms of characteristics and methods of development than those seen in the research environment.

The user interface to a software system can spell the difference between success and failure. Sometimes, function does not seem to count. If the program does a good enough job, if the users see an easy to use, easy to learn, helpful, pleasant interface, they love it. The interface might be the most significant sales aspect of a software product (consider the spate of look-and-feel lawsuits!). This wasn't always the situation.

The goal of this group is to explore the use of domainspecific knowledge and natural deduction-based reasoning techniques to construct theorem provers that operate in nontrivial mathematical domains. Two new provers, by Larry IIines and Tie-Cheng Wang, are very much like expert systems, since the prover takes its direction by trying to satisfy "higher level" goals, based on knowledge about theorem proving. These are stand-alone provers, not man-machine systems, which are attacking some fairly difficult theorems in mathematics. In addition to this mainline work on mathematical theorem provers, two auxiliary efforts rely heavily on knowledge-based deduction. Michael Starbird is developing a knowledge-based expert system for an area of geometric topology, particularly for three dimensions.

Planning domains often feature subproblems such as route planning and resource handling. Using static domain analysis techniques, we have been able to identify certain commonly occurring subproblems within planning domains, making it possible to abstract these subproblems from the overall goals of the planner and deploy specialized technology to handle them in a way integrated with the broader planning activities. Although such strategies can be impressive when applied to toy domains, they cannot address highly structured problem domains effectively. However, when knowledge-sparse approaches are supplemented by domain knowledge, they can perform impressively (Bacchus and Kabanza 2000) at the cost of an increased representation burden on the domain designer.

Michael R. Lowry There is substantial evidence that AI technology can meet the requirements of the large potential market that will exist for knowledge-based software engineering at the turn of the century. In this article, which forms the conclusion to the AAAI Press book Automating Software Design, edited by Michael Lowry and Robert McCartney, Michael Lowry discusses the future of software engineering, and how knowledge-based software engineering (KBSE) progress will lead to system development environments. Specifically, Lowry examines how KBSE techniques promote additive programming methods and how they can be developed and introduced in an evolutionary way. The enabling technology will come from AI, formal methods, programming language theory, and other areas of computer science. This technology will enable much of the knowledge now lost in the software development process to be captured in machineencoded form and automated.

This article discusses t,he selection of the domain for a knowledge-based expert system for a corporate application The selection of the domain is a critical task in an expert system development At the st,art of a project looking into the development of an expert, syst,em, the knowledge engineering project team must investigate one or several possible expert system domains They must decide whether the selected application(s) are best suited to solution by present expert system technology, or if there might he a hettel way (or, possibly, no way) to attack the problems. If there arc several possibilities, the team must also rank the potential applications and select the best availahlc To evaluate the potential of possible application domains, it has proved very useful to have a set of desired at,trihutes for a good expert system domain. This art,iclc presents such a set of attrihut,es The at,trihute set was developed as part of a major expert system development project at GTE Lahorat.ories. In particular, it focuses on selecting an expert system domain for a corporate application. One of the prime arcas of corporate interest is expert systems.

The growth in the amount of available databases far outstrips the growth of corresponding knowledge. This creates both a need and an opportunity for extracting knowledge from databases. Many recent results have been reported on extracting different kinds of knowledge from databases, including diagnostic rules, drug side effects, classes of stars, rules for expert systems, and rules for semantic query optimization. The importance of this topic is now recognized by leading researchers. Michie predicts that "The next area that is going to explode is the use of machine learning tools as a component of large scale data analysis'' (AI Week, March 15, 1990).