Didrik Nielsen, Lin Wu, Mohammad Emtiyaz Khan, Voot Tangkaratt, Wu Lin, Zuozhu Liu
Introduces a variational Newton-based optimization method that adaptively balances exploration and exploitation for improved learning in complex environments.
Open research commons for machine learning
2,250,3162.3MPapersPapers215,519216KAuthorsAuthors28,58829KResearch DirectionsDirections25,12025KProposalsProposals
Updated dailyFollow Lacuna on X
New papers
Show papers for you- HeMeR: Heterogeneous Memory Reconciliation for Embodied Agents via Structured KV Reuse
- BlinkMimic: Training a Humanoid Motion Tracker in Minutes on a Single GPU
- Autonomous Mobile Robots on a Computing Field
- Where in the Input Does a One-Pass Decision Encoder Look?
- InertiaMPDD: Multimodal Personalized Depression Detection via Emotional Inertia-Inspired Dynamics Modeling
Featured Paper
Inducing History Representations for Broad Coverage Statistical Parsing
naacl 2003 120 citations
This paper presents a novel approach to statistical parsing by using Simple Synchrony Networks (SSNs) to automatically induce real-valued representations of unbounded parse histories, as illustrated by the sequential numbering of tree nodes in the provided figure. By moving away from hand-crafted features and rigid independence assumptions, the model learns to compress the entire previous derivation state into a fixed-width vector for decision making. The architecture achieved a state-of-the-art 89.1% F-measure on the Penn Treebank, demonstrating that neural history induction can outperform traditional generative models that rely on limited context windows. (Explore paper...)
Today in ML History — November 15
-
2017
-
2018
Dong Hao, Kai LiDevelops mechanisms to enforce cooperation and prevent collusion among agents in repeated public goods games using game-theoretic strategies.
-
2025
Aditi Bhalla, Christian Hellert, Enkelejda KasneciImplements unsupervised domain adaptation for driver monitoring by aligning features across different camera views and sensor modalities without labels.
Featured Directions
- Factual Inaccuracy and Hallucination in Large Language Models3,605 papers
- Understanding Prompt Sensitivity in Language Models3,146 papers
- Automated Discovery of Data Symmetries471 papers
- Automated Discovery of Optimization Heuristics351 papers
- Temporal Bias and Latency in Citation Networks330 papers
- The Coverage Bias of Single Database Bibliometric Analyses301 papers
Lacuna is a research map of machine learning by Tiptree Systems: papers linked to the research directions, authors, venues and institutions behind them, with summaries that link back to their sources. Free to read, no account needed.
The Lacuna paper is by Martin Weiss, Miles Q. Li, Alejandro H. Artiles, Yacine Mkhinini, Chris Pal, Hugo Larochelle and Nasim Rahaman.