Peer-curated academic archive from Alberta Machine Intelligence Institute (Amii), comprising 5 recorded lectures in AI & Machine Learning, Algorithms & Theoretical CS, Computer Systems & Software.
Self-contained research seminars, public lectures, and invited colloquia.
Introducing Endpoint Replay, an algorithm that significantly compresses recency buffers in deep reinforcement learning by pairing a recency buffer with a coreset of chained n-step transitions.
Exploring techniques for machine learning to automate formal verification, focusing on mining reusable custom tactics and the challenges in applying abstractions to real-world projects.
This seminar presents research on actionable interpretability, translating insights into language model reasoning, learning, and failure modes into concrete improvements in training and inference. Topics include parameter-efficient adaptation via localized computation, distillation dynamics, and gradient-based analysis for reward hacking suppression.
A deep dive into methods for optimizing the Conditional Value-at-Risk (CVaR) in Markov Decision Processes, introducing the Dynamically Augmented CVaR for robust decision-making.
An examination of how large language models bridge lexical and sentence semantics through concepts, presenting recent findings and highlighting limitations in reasoning and multilingual contexts.
Professors, researchers, and invited lecturers recorded at Alberta Machine Intelligence Institute (Amii).
Browse, filter, and sort every lecture from Alberta Machine Intelligence Institute (Amii).
