optimal transport
PulseAugur coverage of optimal transport — every cluster mentioning optimal transport across labs, papers, and developer communities, ranked by signal.
3 day(s) with sentiment data
-
New GPU solver cuRegOT accelerates optimal transport for machine learning
Researchers have developed cuRegOT, a new GPU-accelerated solver designed to overcome the computational challenges of optimal transport (OT) in large-scale machine learning applications. The solver addresses the limitat…
-
Researchers analyze Monge Map stability in semi-dual optimal transport
This paper investigates the semi-dual formulation of optimal transport, revealing its degenerate saddle-point structure. The research establishes conditions for the convergence of Monge maps without assuming dual potent…
-
AI research tackles domain adaptation with new locality-aware private class identification
Researchers have developed a new method called ReOT to improve domain adaptation in machine learning, particularly when dealing with extreme label shifts. This approach uses locality-aware private class identification b…
-
New PLOT framework speeds up neural network interpretability
Researchers have developed PLOT, a new framework for mechanistic interpretability in neural networks. PLOT uses optimal transport to efficiently localize causal variables within a neural network's computation. This meth…
-
New ReshapeOT method improves optimal transport for modeling distribution shifts
Researchers have introduced Displacement-Reshaped Optimal Transport (ReshapeOT), a novel method for modeling distribution shifts. This technique enhances the ground metric used in optimal transport by incorporating obse…
-
New CTAD framework calibrates tabular anomaly detection using optimal transport
Researchers have developed CTAD, a novel post-processing framework designed to enhance the performance of existing tabular anomaly detection methods. CTAD works by characterizing normal data through empirical and struct…
-
New framework unifies entropic OT with neural networks on curved spaces
Researchers have introduced Entropic Riemannian Neural Optimal Transport (Entropic RNOT), a novel framework designed to handle machine learning problems involving data on curved spaces. This method unifies intrinsic ent…
-
New research explores advanced reward modeling for LLMs and diffusion models
Several new research papers explore advancements in reward modeling for AI alignment, particularly for large language models and diffusion models. One paper introduces SelectiveRM, a framework using optimal transport to…
-
New CUDA implementation speeds up optimal transport calculations on GPUs
Researchers have developed FastSinkhorn, a new CUDA implementation for the Sinkhorn algorithm used in optimal transport computations. This method operates entirely in the log-domain, ensuring numerical stability even wi…
-
New methods train neural networks with non-differentiable components
Researchers have developed new methods for training neural networks that incorporate non-differentiable components, a common challenge in areas like spiking neurons or quantized layers. One approach, detailed in an arXi…
-
Aitchison geometry powers new compositional graph embeddings and flavor tagger calibration
Two new arXiv papers introduce novel approaches to representation learning using Aitchison geometry. One paper proposes a framework for calibrating flavor taggers in high-energy physics by formulating it as an optimal t…
-
New USB framework reconstructs discrete cell dynamics from snapshots
Researchers have developed a new simulation-free framework called Unbalanced Schrödinger Bridge (USB) to reconstruct discrete branching dynamics from single-cell data. This method addresses limitations of existing conti…
-
New optimal transport methods offer improved accuracy and scalability
Researchers have introduced Sliced-Regularized Optimal Transport (SROT), a novel formulation that regularizes transport plans towards a smoothed sliced OT plan, offering more accurate approximations than entropic OT. A …
-
New techniques like UniVer and SpecKV boost LLM inference speed via speculative decoding
Researchers have developed new methods to accelerate large language model (LLM) inference. UniVer offers a unified approach to multi-step and multi-draft speculative decoding, improving acceptance length by up to 8.5%. …