Department of Computer Science

Automated Machine Learning (Auto-ML) is a growing field
where several techniques are being developed to address the question of
how to automate the process of defining machine learning pipelines, using
diverse types of approaches and with relative success, but still, being far
from solved. Among these still unsolved questions, the computational
cost is one of the major issues. In this context, evaluating a model takes
a lot of time and resources, and yet that is still a step that has not
received much attention in the Auto-ML literature.
In this sense, this work revisits the Auto-CVE (Automated Coevolutionary Voting Ensemble) and proposes a new method for model evaluation:
the dynamic sampling holdout. When compared to the regular Auto-
CVE with cross-validation and the popular TPOT (Tree-based Pipeline
Optimization Tool) algorithm, Auto-CVE with dynamic holdout shows
competitive results in both predictive performance and computing time.

Ontologies are instruments of knowledge organization that have been developed by using several methodologies. These methodologies are well-known and established, but their steps often are not well explained. Thus, only knowledge engineers, in general with a major in computer science, are able to perform all steps required in ontologies development. In the present paper, we describe a methodology that details each step of the ontology development cycle. The goal of this methodology, which is called OntoForInfoScience, is to overcome issues that arise from the use of technical jargon and from the application of logical-philosophical principles in the ontology development. In order to identify the afore-mentioned issues, our methodology was produced by information scientists during the development of ontologies in the domain of human blood. Such domain ontology, called Hem-onto, is part of an ongoing scientific biomedical project. In this paper, we present a brief description of OntoForInfoSci-ence, as well as some results of its application in the development of the blood ontology. Finally, we conclude that the new methodology is useful for information scientists creating formal ontological representations. In addition to the methodology per se, we also provide partial results of Hemonto development.