Graduate Theses: Recent submissions
Now showing items 529-531 of 36544
-
Switching State Space Modeling via Constrained Inference for Clinical Outcome Prediction
(Massachusetts Institute of Technology, 2025-05)In clinical settings, timely and accurate prediction of adverse patient outcomes can help guide treatment decisions. While deep learning models such as LSTMs have demonstrated strong predictive performance on multivariate ... -
Duality, Weight Decay, and Metrized Deep Learning
(Massachusetts Institute of Technology, 2025-05)The Muon optimizer has shown convincing evidence that it is faster and more scalable than AdamW for deep learning training, setting speed records for training NanoGPT and scaling up to models with 16B parameters. The theory ... -
Modeling Sequence Uncertainty in Comparative Genomics with a Probabilistic DNA Representation
(Massachusetts Institute of Technology, 2025-05)Uncertainty in nucleotide sequences is widespread in bioinformatics, arising from somatic mutations, population-level variation, sequencing errors, and ancestral state inference. Yet, standard formats like FASTA encode DNA ...
