ICAL Mission Statement
The IC@L mission is to support researchers at the Lassonde School of Engineering and other participating faculties at York University in exploring innovative computational methodologies. These efforts focus on advancing quantitative, qualitative, interdisciplinary, and experimental approaches to design next-generation computing systems that are compliant, efficient, safe, secure, adaptive, and intelligent.
Through these efforts, IC@L seeks to address critical societal challenges across a wide range of domains, including environmental sustainability, smart cities, healthcare, education, finance, and governance. By fostering innovation and collaboration, IC@L is committed to building the future of computing to create transformative impacts on society.
Featured Papers
-
Goldman, J., & Tsotsos, J. K. (2024). Statistical challenges with dataset construction: Why you will never have enough images.
Published on: arXiv preprint arXiv:2408.11160
Deep neural networks perform well on benchmarks, but this doesn’t guarantee success in real-world, safety-critical environments. Testing with representative datasets is often impractical, and performance metrics based on non-representative data are unreliable. The paper suggests shifting from accuracy-focused evaluations to assessing a model’s decision-making process, as current test set-based methods lack statistical reliability.
Member’s Spotlight
-
Goldman, J., & Tsotsos, J. K. (2024). Statistical challenges with dataset construction: Why you will never have enough images.
Published on: arXiv preprint arXiv:2408.11160
Deep neural networks perform well on benchmarks, but this doesn’t guarantee success in real-world, safety-critical environments. Testing with representative datasets is often impractical, and performance metrics based on non-representative data are unreliable. The paper suggests shifting from accuracy-focused evaluations to assessing a model’s decision-making process, as current test set-based methods lack statistical reliability.