Université de Montréal

About: Université de Montréal is a education organization based out in Montreal, Quebec, Canada. It is known for research contribution in the topics: Population & Context (language use). The organization has 45641 authors who have published 100476 publications receiving 4004007 citations. The organization is also known as: University of Montreal & UdeM.

On the Spectral Bias of Neural Networks

Abstract: Neural networks are known to be a class of highly expressive functions able to fit even random input-output mappings with $100\%$ accuracy. In this work, we present properties of neural networks that complement this aspect of expressivity. By using tools from Fourier analysis, we show that deep ReLU networks are biased towards low frequency functions, meaning that they cannot have local fluctuations without affecting their global behavior. Intuitively, this property is in line with the observation that over-parameterized networks find simple patterns that generalize across data samples. We also investigate how the shape of the data manifold affects expressivity by showing evidence that learning high frequencies gets \emph{easier} with increasing manifold complexity, and present a theoretical understanding of this behavior. Finally, we study the robustness of the frequency components with respect to parameter perturbation, to develop the intuition that the parameters must be finely tuned to express high frequency functions.

Nomograms for predicting survival of patients with newly diagnosed glioblastoma: prognostic factor analysis of EORTC and NCIC trial 26981-22981/CE.3

Abstract: Summary Background A randomised trial published by the European Organisation for Research and Treatment of Cancer (EORTC) and the National Cancer Institute of Canada (NCIC) Clinical Trials Group (trial 26981-22981/CE.3) showed that addition of temozolomide to radiotherapy in the treatment of patients with newly diagnosed glioblastoma significantly improved survival. We aimed to undertake an exploratory subanalysis of the EORTC and NCIC data to confirm or identify new prognostic factors for survival in adult patients with glioblastoma, derive nomograms that predict an individual patient's prognosis, and suggest stratification factors for future trials. Methods Data from 573 patients with newly diagnosed glioblastoma who were randomly assigned to radiotherapy alone or to the same radiotherapy plus temozolomide in the EORTC and NCIC trial were included in this subanalysis. Survival modelling was done in three patient populations: intention-to-treat population of all randomised patients (population 1); patients assigned temozolomide and radiotherapy (population 2, n=287); and patients assigned temozolomide and radiotherapy who had assessment of MGMT promoter methylation status and who had undergone tumour resection (population 3, n=103). Cox proportional hazards models were fitted with and without O6-methylguanine-DNA methyltransferase ( MGMT ) promoter methylation status. Nomograms were developed to predict an individual patient's median and 2-year survival probabilities. No nomogram was developed in the radiotherapy-alone group because combined treatment is now the new standard of care. Findings Independent of the MGMT promoter methylation status, analysis in all randomised patients (population 1) identified combined treatment with temozolomide, more extensive tumour resection, younger age, Mini-Mental State Examination (MMSE) score of 27 or higher, and no corticosteroid treatment at baseline as independent prognostic factors correlated with improved survival outcome. In patients assigned temozolomide and radiotherapy (population 2), younger age, better performance status, more extensive tumour resection, and MMSE score of 27 or higher were associated with better survival. In patients who had tumours resected, who were assigned temozolomide and radiotherapy, and who had available MGMT promoter methylation status (population 3), methylated MGMT , better performance status, and MMSE score of 27 or higher were associated with improved survival. Nomograms were developed and are available at http://www.eortc.be/tools/gbmcalculator. Interpretation MGMT promoter methylation status, age, performance status, extent of resection, and MMSE are suggested as eligibility or stratification factors for future trials in patients with newly diagnosed glioblastoma. Stratifying by MGMT promoter methylation status should be mandatory in all glioblastoma trials that use alkylating chemotherapy. Nomograms can be used to predict an individual patient's prognosis, and they integrate pertinent molecular information that is consistent with a paradigm shift towards individualised patient management.

Barriers to physical activity among patients with type 1 diabetes.

Abstract: OBJECTIVE—To determine, in an adult population with type 1 diabetes, barriers to regular physical activity using a diabetes-specific barriers measure (the Barriers to Physical Activity in Diabetes [type 1] [BAPAD1] scale) and factors associated with these barriers. RESEARCH DESIGN AND METHODS—One hundred adults with type 1 diabetes answered a questionnaire assessing perceived barriers to physical activity and related factors. A1C was obtained from the medical chart of each individual. RESULTS—Fear of hypoglycemia was identified as being the strongest barrier to physical activity. Greater knowledge about insulin pharmacokinetics and using appropriate approaches to minimize exercise-induced hypoglycemia were factors associated with fewer perceived barriers. Greater barriers were positively correlated with A1C levels (r = 0.203; P = 0.042) and negatively with well-being (r = −0.45; P < 0.001). CONCLUSIONS—Fear of hypoglycemia is the strongest barrier to regular physical activity in adults with type 1 diabetes, who should therefore be informed and supported in hypoglycemia management.