A few of the ideas, topics, and commonplaces that have been gaining steam on arXiv during the last few months (explainer)
1. Flow prediction: Two different meanings here: flows in the sense of Bayesian normalizing flows (a useful method to sample from complex distributions through simpler ones), and flows in the sense of material flow analyses (an approach to study, well, the flow of material in some process or system). Both interesting!
Some recent articles:
- Emulating compact binary population synthesis simulations with robust uncertainty quantification and model comparison: Bayesian normalizing flows
- Accelerated Bayesian calibration and uncertainty quantification of RANS turbulence model parameters for stratified atmospheric boundary layer flows
- Bayesian Model Selection for Network Discrimination and Risk-informed Decision Making in Material Flow Analysis
- Intelligent data collection for network discrimination in material flow analysis using Bayesian optimal experimental design
2. Institutional design: I'm partial to this problem: the structure and culture of an organization will determine its cognitive capabilities at least as much as the other way around. Amateurs study tactics, etc.
Some recent articles:
- Refugees' path to legal stability is long and systematically unequal
- Finance as Extended Biology: Reciprocity as the Cognitive Substrate of Financial Behavior
- Selecting Optimal Candidate Profiles in Adversarial Environments Using Conjoint Analysis and Machine Learning
- Uncertainty, bias and the institution bootstrapping problem
3. Memory retrieval: There are more neural networks here than I'd like — after all the idea of these posts is to highlight things outside generative AI — but deeper (no pun intended) looks at how memory works there and elsewhere is fascinating on its own. In a way, the more we understand how a system processes information the less likely we are to project on them our own skeuomorphic assumptions.
Some recent articles:
- Nonparametric Modern Hopfield Models
- Transient dynamics of associative memory models
- Memorization to Generalization: Emergence of Diffusion Models from Associative Memory
- MIRROR: Cognitive Inner Monologue Between Conversational Turns for Persistent Reflection and Reasoning in Conversational LLMs
- CAIM: Development and Evaluation of a Cognitive AI Memory Framework for Long-Term Interaction with Intelligent Agents