Ben Recht is a researcher and writer exploring the history, theory, and practice of decision-making by humans and machines. On arg min, he covers optimization, machine learning, cybernetics, and occasional reflections on music and culture.
16 articles from this blog
A lecture blog exploring surprising robustness and fragility in interconnected systems, drawing analogies between nonlinear control and optimization.
A lecture on control theory, explaining feedback as an algebraic interconnection for robust system design, using amplifier gain as an example.
A syllabus for a university course exploring the principles of feedback, control theory, and their connections to machine learning and optimization.
A professor reflects on the intersection of machine learning and control theory, discussing the Learning for Dynamics and Control (L4DC) conference and the need for a merged perspective.
A blog post arguing that statistical inference is often used as a tool of rhetoric and persuasion, rather than pure objective science.
A critique of statistical inference's reliance on p-values and combinatorics, arguing it obscures real-world causality and individual context.
A year-in-review blog post reflecting on machine learning course blogging, revisiting 'The Bitter Lesson', and critiquing trends in ML and economics.
Critique of causal inference in statistics, highlighting the flawed assumption that treatments have no impact on future outcomes, using cancer screening trials as an example.
Explores the tension between optimization and systems-level thinking in AI-driven scientific discovery and computational ethics.
A critique of using AI to automate science, arguing that metrics have become goals, distorting scientific progress.
A professor shares open research problems inspired by his graduate machine learning class, focusing on design-based ML and competitive testing theory.
A lecture reflection on the gap between mathematical theory and practical engineering in machine learning, arguing for social analysis over functional analysis.
A machine learning professor critiques the foundational concept of a 'data-generating distribution' and shares insights from teaching a truly distribution-free course.
A critique of Reformist RL's inefficiency and a proposal for more effective alternatives in reinforcement learning.
A simplified, non-technical definition of reinforcement learning as an iterative optimization process based on external feedback.
A technical lecture on applying policy gradient methods to derive optimization algorithms, focusing on the unbiased gradient estimator and its applications.
