paper-with-me

홈 › Papers

The Difficulty of Training Sparse Neural Networks

2019-06-25 · ICML Workshop Deep_Phenomen 2019 6 · Utku Evci, Fabian Pedregosa, Aidan Gomez, Erich Elsen

We investigate the difficulties of training sparse neural networks and make new observations about optimization dynamics and the energy landscape within the sparse regime. Recent work of \citep{Gale2019, Liu2018} has shown that sparse ResNet-50 architectures trained on ImageNet-2012 dataset converge to solutions that are significantly worse than those found by pruning. We show that, despite the failure of optimizers, there is a linear path with a monotonically decreasing objective from the initialization to the "good" solution. Additionally, our attempts to find a decreasing objective path from "bad" solutions to the "good" ones in the sparse subspace fail. However, if we allow the path to traverse the dense subspace, then we consistently find a path between two solutions. These findings suggest traversing extra dimensions may be needed to escape stationary points found in the sparse subspace.

📄 PDF Abstract BibTeX arXiv:1906.10732

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Goldilocks RL: Tuning Task Difficulty to Escape Sparse Rewards for Reasoning

2026-02-16 · Ilia Mahrooghi, Aryo Lotfi, Emmanuel Abbe arxiv

Reinforcement learning has emerged as a powerful paradigm for unlocking reasoning capabilities in language models. However, relying on sparse rewards makes this process highly sample-inefficient, as models must navigate …

Reinforcement Learning

Dual-Criterion Curriculum Learning: Application to Temporal Data

2026-03-24 · Gaspard Abel, Eloi Campagne, Mohamed Benloughmari, Argyris Kalogeratos arxiv

Curriculum Learning (CL) is a meta-learning paradigm that trains a model by feeding the data instances incrementally according to a schedule, which is based on difficulty progression. Defining meaningful difficulty asses…

DIVA-GRPO: Enhancing Multimodal Reasoning through Difficulty-Adaptive Variant Advantage

2026-03-01 · Haowen Gao, Zhenyu Zhang, Liang Pang, Fangda Guo 외 arxiv

Reinforcement learning (RL) with group relative policy optimization (GRPO) has become a widely adopted approach for enhancing the reasoning capabilities of multimodal large language models (MLLMs). While GRPO enables lon…

Reinforcement LearningMultimodal Reasoning

Accurate Neural Network Pruning Requires Rethinking Sparse Optimization

2023-08-03 · Denis Kuznedelev, Eldar Kurtic, Eugenia Iofinova, Elias Frantar 외

Obtaining versions of deep neural networks that are both highly-accurate and highly-sparse is one of the main challenges in the area of model compression, and several high-performance pruning techniques have been investi…

Model CompressionNetwork PruningStochastic Optimization

Guiding LLM Post-training Data Engineering with Model Internals from Sparse Autoencoders

2026-05-26 · Yi Jing, Zao Dai, Jinwu Hu, Zijun Yao 외 arxiv

Model internals encode rich information about how a large language model (LLM) processes its training data; however, post-training data engineering largely relies on external signals and ignores rich intrinsic signals ly…

Reinforcement Learning