We consider the problem of learning multiple tasks in a continual learning setting in which data from different tasks is presented to the learner in a streaming fashion. A key challenge in this... Show moreWe consider the problem of learning multiple tasks in a continual learning setting in which data from different tasks is presented to the learner in a streaming fashion. A key challenge in this setting is the so-called "catastrophic forgetting problem", in which the performance of the learner in an "old task" decreases when subsequently trained on a "new task". Existing continual learning methods, such as Averaged Gradient Episodic Memory (A-GEM) and Orthogonal Gradient Descent (OGD), address catastrophic forgetting by minimizing the loss for the current task without increasing the loss for previous tasks. However, these methods assume the learner knows when the task changes, which is unrealistic in practice. In this paper, we alleviate the need to provide the algorithm with information about task changes by using an online clustering-based approach on a dynamically updated finite pool of samples or gradients. We thereby successfully counteract catastrophic forgetting in one of the hardest settings, namely: domain-incremental learning, a setting for which the problem was previously unsolved. We showcase the benefits of our approach by applying these ideas to projection-based methods, such as A-GEM and OGD, which lead to task-agnostic versions of them. Experiments on real datasets demonstrate the effectiveness of the proposed strategy and its promising performance compared to state-of-the-art methods. Show less
Lamers, C.H.C.B.; Vidal, R.; Belbachir, N.; Stein, N. van; Bäck, T.H.W.; Paris, G. 2023
Background: Individuals with a non-syndromic family history of colorectal cancer are known to have an increased risk. There is an opportunity to prevent early-onset colorectal cancer (age less than... Show moreBackground: Individuals with a non-syndromic family history of colorectal cancer are known to have an increased risk. There is an opportunity to prevent early-onset colorectal cancer (age less than 50 years) (EOCRC) in this population. The aim was to explore the proportion of EOCRC that is preventable due to family history of colorectal cancer. Methods: This was a retrospective multicentre European study of patients with non-hereditary EOCRC. The impact of the European Society of Gastrointestinal Endoscopy (ESGE), U.S. Multi-Society Task Force (USMSTF), and National Comprehensive Cancer Network (NCCN) guidelines on prevention and early diagnosis was compared. Colorectal cancer was defined as potentially preventable if surveillance colonoscopy would have been performed at least 5 years before the age of diagnosis of colorectal cancer, and diagnosed early if colonoscopy was undertaken between 1 and 4 years before the diagnosis. Results: Some 903 patients with EOCRC were included. Criteria for familial colorectal cancer risk in ESGE, USMSTF, and NCCN guidelines were met in 6.3, 9.4, and 30.4 per cent of patients respectively. Based on ESGE, USMSTF, and NCCN guidelines, colorectal cancer could potentially have been prevented in 41, 55, and 30.3 per cent of patients, and diagnosed earlier in 11, 14, and 21.1 per cent respectively. In ESGE guidelines, if surveillance had started 10 years before the youngest relative, there would be a significant increase in prevention (41 versus 55 per cent; P = 0.010). Conclusion: ESGE, USMSTF, and NCCN criteria for familial colorectal cancer were met in 6.3, 9.4, and 30.4 per cent of patients with EOCRC respectively. In these patients, early detection and/or prevention could be achieved in 52, 70, and 51.4 per cent respectively. Early and accurate identification of familial colorectal cancer risk and increase in the uptake of early colonoscopy are key to decreasing familial EOCRC. Show less
