paper-with-me

홈 › Papers

Loss Surfaces, Mode Connectivity, and Fast Ensembling of DNNs

2018-02-27 · NeurIPS 2018 12 · Timur Garipov, Pavel Izmailov, Dmitrii Podoprikhin, Dmitry Vetrov, Andrew Gordon Wilson

The loss functions of deep neural networks are complex and their geometric properties are not well understood. We show that the optima of these complex loss functions are in fact connected by simple curves over which training and test accuracy are nearly constant. We introduce a training procedure to discover these high-accuracy pathways between modes. Inspired by this new geometric insight, we also propose a new ensembling method entitled Fast Geometric Ensembling (FGE). Using FGE we can train high-performing ensembles in the time required to train a single model. We achieve improved performance compared to the recent state-of-the-art Snapshot Ensembles, on CIFAR-10, CIFAR-100, and ImageNet.

📄 PDF Abstract BibTeX arXiv:1802.10026

Code (8)

timgaripov/dnn-mode-connectivity 공식 구현 pytorch
DeBur19/FGE_reproduction_project pytorch
biomedia-mbzuai/fissionfusion pytorch
chandansharma02/Deep_Learning pytorch
g-benton/loss-surface-simplexes pytorch
simon-larsson/keras-swa tf
tjwhitaker/prune-and-tune-ensembles pytorch
xuyxu/Ensemble-Pytorch pytorch

Methods 이 논문이 사용한 방법론

Snapshot Ensembles The overhead cost of training multiple deep neural networks could be very high in terms of the training time, hardware, and computational resource requirement and…

Similar Papers 제목 키워드 기반

Loss Surface Simplexes for Mode Connecting Volumes and Fast Ensembling

2021-02-25 · Gregory W. Benton, Wesley J. Maddox, Sanae Lotfi, Andrew Gordon Wilson

With a better understanding of the loss surfaces for multilayer networks, we can build more robust and accurate training procedures. Recently it was discovered that independently trained SGD solutions can be connected al…

Traversing Between Modes in Function Space for Fast Ensembling

2023-06-20 · Eunggu Yun, Hyungi Lee, Giung Nam, Juho Lee

Deep ensemble is a simple yet powerful way to improve the performance of deep neural networks. Under this motivation, recent works on mode connectivity have shown that parameters of ensembles are connected by low-loss su…

Using Mode Connectivity for Loss Landscape Analysis

2018-06-18 · Akhilesh Gotmare, Nitish Shirish Keskar, Caiming Xiong, Richard Socher

Mode connectivity is a recently introduced frame- work that empirically establishes the connected- ness of minima by finding a high accuracy curve between two independently trained models. To investigate the limits of th…

A Generative Model for Sampling High-Performance and Diverse Weights for Neural Networks

2019-05-07 · Lior Deutsch, Erik Nijkamp, Yu Yang

Recent work on mode connectivity in the loss landscape of deep neural networks has demonstrated that the locus of (sub-)optimal weight vectors lies on continuous paths. In this work, we train a neural network that serves…

Computational EfficiencyDiversity

When to Ensemble: Identifying Token-Level Points for Stable and Fast LLM Ensembling

2025-10-17 · Heecheol Yun, Kwangmin Ki, Junghyun Lee, Eunho Yang arxiv

Ensembling Large Language Models (LLMs) has gained attention as a promising approach to surpass the performance of individual models by leveraging their complementary strengths. In particular, aggregating models' next-to…