Sheharyar Raza: Managing Randomness in Artificial Intelligence for Transfusion Medicine
Sheharyar Raza, Transfusion Medicine Specialist at Sunnybrook Health Sciences Center and Internal Medicine Physician at Unity Health Toronto, shared a post on LinkedIn about a recent article he and Ruchika Goel co-authored, published in Vox Sanguinis, adding:
“Brief response to a letter published about our recent article by Ruchika Goel and I.
Our core point is that the distinction between newer generative models and ‘traditional’ machine learning models isn’t that one is stochastic and one is deterministic.
Depending on your specific use and pipeline, a language model architecture can be effectively deterministic (e.g. embedding models), while a traditional ML model can behave highly non-deterministically (e.g. clustering with a non-convex objective or bootstrap-based methods with small n).
Methods for evaluating the two are also roughly analogous, even if they appear superficially different.
What matters is to look at steps where, and how much, randomness comes into your model, and how it is being handled.
Thanks Aman for helpful discussion.”
Title: Response to comment on ‘Harnessing big data and artificial intelligence in transfusion medicine: Opportunities for precision, safety and efficiency’
Authors: Sheharyar Raza, Ruchika Goel
Other posts featuring Sheharyar Raza on Hemostasis Today.
-
Sep 15, 2026, 17:33Reza Shojaei: Blood and Plasma Donors Are Not a Production Metric
-
Sep 15, 2026, 17:32Thomas Dayspring: MHT for Vasomotor Symptoms May Also Offer Cardiovascular Benefits
-
Sep 15, 2026, 17:31Rebecca Risman: Deciphering the Mechanisms of Fibrinolysis in Contracted Clots
-
Sep 15, 2026, 17:31Clare Tower: 30 Times Higher Does Not Tell the Whole Story
-
Sep 15, 2026, 17:29Erica Wood: Looking Forward to the Webinar on Future Directions in Haemovigilance
-
Sep 15, 2026, 17:28Caitlin Raymond: Can an Antibody Screen Tell You If a Transfusion Will Actually Hurt Your Patient?
-
Sep 15, 2026, 17:28Rachel Margolis-Goodman: When Conventional Treatment Could Not, Research Found an Answer
-
Sep 15, 2026, 17:27Carly Newton: Closing the Gap in Sickle Cell Disease Readmissions
-
Sep 15, 2026, 17:26Jared Rice: Understanding Extreme Platelet Counts in Preterm Infants