搜索结果：Mona

The Muon optimizer has recently offered a promising alternative to AdamW for large language model training, leveraging matrix orthogonalization to produce geometry-aware updates. However, like all first-order methods, Muon can become trapped in sharp local minima. In this work, we present MONA, an optimizer that bridges Muon's orthogonalization framework with curvature-aware acceleration. MONA adds an acceleration term directly into Muon's gradient processing pipeline. This term is calculated from the exponential moving average of gradient differences. We provide a detailed convergence analysis for MONA, showing that the acceleration term enables escape from sharp minima while preserving Muon's spectral-norm regularization. Empirically, MONA achieves better convergence and downstream task performance compared to both Muon and AdamW across three scales of Mixture-of-Experts pretraining, spanning from 1B to 68B parameters, with the largest model trained on 1 trillion tokens. Furthermore, we conduct supervised fine-tuning on the MOE-68B-A3B model and evaluate it on general capability, mathematical reasoning, and code generation benchmarks, where MONA achieves SOTA performance.

Reasoning about consistency models for replicated data systems is a challenging task that requires a deep understanding of both the consistency models themselves and a large part of human inputs in mechanized verification approaches. In this work, we introduce an approach to reasoning about consistency models for replicated data systems. We introduce HistMSO, a monadic second-order logic (MSO) for histories and abstract executions, the formal models of executions of replicated data systems introduced by Burckhardt. We show that HistMSO can express 39 out of 42 consistency models from Viotti and Vukolic hierarchy. Moreover, we develop a method for reducing HistMSO satisfiability and model-checking to the same problems for MSO over words. While doing this, we leverage the MONA tool for automated reasoning on consistency models.

Future advanced AI systems may learn sophisticated strategies through reinforcement learning (RL) that humans cannot understand well enough to safely evaluate. We propose a training method which avoids agents learning undesired multi-step plans that receive high reward (multi-step "reward hacks") even if humans are not able to detect that the behaviour is undesired. The method, Myopic Optimization with Non-myopic Approval (MONA), works by combining short-sighted optimization with far-sighted reward. We demonstrate that MONA can prevent multi-step reward hacking that ordinary RL causes, even without being able to detect the reward hacking and without any extra information that ordinary RL does not get access to. We study MONA empirically in three settings which model different misalignment failure modes including 2-step environments with LLMs representing delegated oversight and encoded reasoning and longer-horizon gridworld environments representing sensor tampering.

Dynamic urban environments, characterized by moving cameras and objects, pose significant challenges for camera trajectory estimation by complicating the distinction between camera-induced and object motion. We introduce MONA, a novel framework designed for robust moving object detection and segmentation from videos shot by dynamic cameras. MONA comprises two key modules: Dynamic Points Extraction, which leverages optical flow and tracking any point to identify dynamic points, and Moving Object Segmentation, which employs adaptive bounding box filtering, and the Segment Anything for precise moving object segmentation. We validate MONA by integrating with the camera trajectory estimation method LEAP-VO, and it achieves state-of-the-art results on the MPI Sintel dataset comparing to existing methods. These results demonstrate MONA's effectiveness for moving object detection and its potential in many other applications in the urban planning field.

Neural Architecture Search (NAS) has proven effective in discovering new Convolutional Neural Network (CNN) architectures, particularly for scenarios with well-defined accuracy optimization goals. However, previous approaches often involve time-consuming training on super networks or intensive architecture sampling and evaluations. Although various zero-cost proxies correlated with CNN model accuracy have been proposed for efficient architecture search without training, their lack of hardware consideration makes it challenging to target highly resource-constrained edge devices such as microcontroller units (MCUs). To address these challenges, we introduce MONAS, a novel hardware-aware zero-shot NAS framework specifically designed for MCUs in edge computing. MONAS incorporates hardware optimality considerations into the search process through our proposed MCU hardware latency estimation model. By combining this with specialized performance indicators (proxies), MONAS identifies optimal neural architectures without incurring heavy training and evaluation costs, optimizing for both hardware latency and accuracy under resource constraints. MONAS achieves up to a 1104x improvement in se

We present MONAS (MicrOdosimetry-based modelliNg for relative biological effectiveness (RBE) ASsessment) toolkit. MONAS is a TOPAS Monte Carlo extension, that combines simulations of microdosimetric distributions with radiobiological microdosimetry-based models for predicting cell survival curves and dose-dependent RBE. MONAS expands TOPAS microdosimetric extension, by including novel specific energy scorers. These spectra are used as physical input to three different formulations of the Microdosimetric Kinetic Model (MKM), and to the Generalized Stochastic Microdosimetric Model (GSM2), to predict dose-dependent cell survival fraction and RBE. MONAS predictions are then validated against experimental microdosimetric spectra and in vitro survival fraction data. We present two different applications of the code: i) the depth-RBE curve calculation from a passively scattered proton SOBP, and ii) the calculation of the 3D RBE distribution on a real head and neck patient geometry treated with protons. MONAS can estimate dose dependent RBE and cell survival curves from experimentally validated microdosimetric spectra with four clinically relevant radiobiological models. From the radiobiol

The Modular Neutron Array (MoNA) collaboration was initiated in 2000 at the National Superconducting Cyclotron Laboratory (NSCL) at Michigan State University (MSU). Since then, the collaboration studied properties of nuclides at and beyond the neutron dripline discovering seven new isotopes between lithium to fluorine. The collaboration included liberal arts colleges, regional comprehensive universities, and major research universities with the focus of giving undergraduate students meaningful research experiences. Over the last 25 years, the combined efforts of hundreds of undergraduates, dozens of graduate students and research associates, and faculty from more than a dozen colleges and universities produced over fifty publications, won awards for research, and combined research and teaching in new and interesting ways.

The concept of k-core plays an important role in measuring the cohesiveness and engagement of a network. And recent studies have shown the vulnerability of k-core under adversarial attacks. However, there are few researchers concentrating on the vulnerability of individual nodes within k-core. Therefore, in this paper, we attempt to study Targeted k-Nodes Collapse Problem (TNsCP), which focuses on removing a minimal size set of edges to make multiple target k-nodes collapse. For this purpose, we first propose a novel algorithm named MOD for candidate reduction. Then we introduce an efficient strategy named MONA, based on MOD, to address TNsCP. Extensive experiments validate the effectiveness and scalability of MONA compared to several baselines. An open-source implementation is available at https://github.com/Yocenly/MONA.

The portrait of Mona Lisa is scrutinized with reference to quantum mechanics. The elements of different expressions are firstly recognized on her face. The contradictory details are then classified in two pictures that, undoubtedly representing distinct moods, confirm dichotomous character of the original. Consecutive discussion has lead to conclusion that the mysterious state Mona Lisa is in actually is coherent mixture - superposition, of cheerfulness and sadness.

Recent studies on neural architecture search have shown that automatically designed neural networks perform as good as expert-crafted architectures. While most existing works aim at finding architectures that optimize the prediction accuracy, these architectures may have complexity and is therefore not suitable being deployed on certain computing environment (e.g., with limited power budgets). We propose MONAS, a framework for Multi-Objective Neural Architectural Search that employs reward functions considering both prediction accuracy and other important objectives (e.g., power consumption) when searching for neural network architectures. Experimental results showed that, compared to the state-ofthe-arts, models found by MONAS achieve comparable or better classification accuracy on computer vision applications, while satisfying the additional objectives such as peak power.

A painting consists of objects which are arranged in specific ways. The art of painting is drawing the objects, which can be considered as known trends, in an expressive manner. Detrended methods are suitable for characterizing the artistic works of the painter by eliminating trends. It means that we study the paintings, regardless of its apparent purpose, as a stochastic process. We apply multifractal detrended fluctuation analysis to characterize the statistical properties of Mona Lisa, as an instance, to exhibit the fractality of the painting. Our results show that Mona Lisa is long range correlated and almost behaves similar in various scales.