English
Related papers

Related papers: Mimetic Initialization of Self-Attention Layers

200 papers

The proper initialization of weights is crucial for the effective training and fast convergence of deep neural networks (DNNs). Prior work in this area has mostly focused on balancing the variance among weights per layer to maintain…

Machine Learning · Computer Science 2020-06-05 Maciej Skorski , Alessandro Temperoni , Martin Theobald

Vision Transformers (ViTs) is emerging as an alternative to convolutional neural networks (CNNs) for visual recognition. They achieve competitive results with CNNs but the lack of the typical convolutional inductive bias makes them more…

Computer Vision and Pattern Recognition · Computer Science 2022-01-27 Yun-Hao Cao , Hao Yu , Jianxin Wu

Layer-sequential unit-variance (LSUV) initialization - a simple method for weight initialization for deep net learning - is proposed. The method consists of the two steps. First, pre-initialize weights of each convolution or inner-product…

Machine Learning · Computer Science 2016-02-22 Dmytro Mishkin , Jiri Matas

Pre-training is prevalent in nowadays deep learning to improve the learned model's performance. However, in the literature on federated learning (FL), neural networks are mostly initialized with random weights. These attract our interest in…

Machine Learning · Computer Science 2023-03-24 Hong-You Chen , Cheng-Hao Tu , Ziwei Li , Han-Wei Shen , Wei-Lun Chao

Stable and efficient training of ReLU networks with large depth is highly sensitive to weight initialization. Improper initialization can cause permanent neuron inactivation dying ReLU and exacerbate gradient instability as network depth…

Machine Learning · Computer Science 2025-09-03 Hyungu Lee , Taehyeong Kim , Hayoung Choi

Although transformer networks are recently employed in various vision tasks with outperforming performance, extensive training data and a lengthy training time are required to train a model to disregard an inductive bias. Using trainable…

Computer Vision and Pattern Recognition · Computer Science 2022-11-28 Heegon Jin , Jongwon Choi

Deeper Vision Transformers (ViTs) are more challenging to train. We expose a degradation problem in deeper layers of ViT when using masked image modeling (MIM) for pre-training. To ease the training of deeper ViTs, we introduce a…

Computer Vision and Pattern Recognition · Computer Science 2023-11-16 Guoxi Huang , Hongtao Fu , Adrian G. Bors

Diffusion Transformers (DiT) have become a leading architecture in image generation. However, the quadratic complexity of attention mechanisms, which are responsible for modeling token-wise relationships, results in significant latency when…

Computer Vision and Pattern Recognition · Computer Science 2024-12-23 Songhua Liu , Zhenxiong Tan , Xinchao Wang

We report competitive results on object detection and instance segmentation on the COCO dataset using standard models trained from random initialization. The results are no worse than their ImageNet pre-training counterparts even when using…

Computer Vision and Pattern Recognition · Computer Science 2018-11-22 Kaiming He , Ross Girshick , Piotr Dollár

Self-training is a simple semi-supervised learning approach: Unlabelled examples that attract high-confidence predictions are labelled with their predictions and added to the training set, with this process being repeated multiple times.…

Computer Vision and Pattern Recognition · Computer Science 2021-09-13 Attaullah Sahito , Eibe Frank , Bernhard Pfahringer

In practice, we usually need to build variable-sized models adapting for diverse resource constraints in different application scenarios, where weight initialization is an important step prior to training. The Learngene framework,…

Machine Learning · Computer Science 2024-04-29 Shi-Yu Xia , Wenxuan Zhu , Xu Yang , Xin Geng

Visual recognition algorithms are required today to exhibit adaptive abilities. Given a deep model trained on a specific, given task, it would be highly desirable to be able to adapt incrementally to new tasks, preserving scalability as the…

Computer Vision and Pattern Recognition · Computer Science 2018-06-15 Massimiliano Mancini , Elisa Ricci , Barbara Caputo , Samuel Rota Bulò

Deploying large-scale transformer models on edge devices presents significant challenges due to strict constraints on memory, compute, and latency. In this work, we propose a lightweight yet effective multi-stage optimization pipeline…

Machine Learning · Computer Science 2025-12-24 Shoaib Mohammad , Guanqun Song , Ting Zhu

Neural Networks are function approximators that have achieved state-of-the-art accuracy in numerous machine learning tasks. In spite of their great success in terms of accuracy, their large training time makes it difficult to use them for…

Machine Learning · Computer Science 2017-04-18 Abhishek Sinha , Mausoom Sarkar , Aahitagni Mukherjee , Balaji Krishnamurthy

Input-Convex Neural Networks (ICNNs) are networks that guarantee convexity in their input-output mapping. These networks have been successfully applied for energy-based modelling, optimal transport problems and learning invariances. The…

Machine Learning · Computer Science 2023-12-21 Pieter-Jan Hoedt , Günter Klambauer

Large Vision Language Models show impressive performance across image and video understanding tasks, yet their computational cost grows rapidly with the number of visual tokens. Existing token pruning methods mitigate this issue through…

Computer Vision and Pattern Recognition · Computer Science 2026-04-02 Dong-Jae Lee , Sunghyun Baek , Junmo Kim

Transformer-based architectures have become a dominant paradigm in vision and language, but their success is often attributed to large model capacity and massive training data. In this work, we examine how self-supervised pre-training,…

Computer Vision and Pattern Recognition · Computer Science 2026-01-06 Oli Bridge , Huey Sun , Botond Branyicskai-Nagy , Charles D'Ornano , Shomit Basu

Fine-tuned transformer models have shown superior performances in many natural language tasks. However, the large model size prohibits deploying high-performance transformer models on resource-constrained devices. This paper proposes a…

Computation and Language · Computer Science 2024-10-01 Zi Yang , Samridhi Choudhary , Siegfried Kunzmann , Zheng Zhang

Masked image modeling (MIM) performs strongly in pre-training large vision Transformers (ViTs). However, small models that are critical for real-world applications cannot or only marginally benefit from this pre-training approach. In this…

Computer Vision and Pattern Recognition · Computer Science 2023-01-04 Sucheng Ren , Fangyun Wei , Zheng Zhang , Han Hu

The training of diffusion models is computationally intensive, making effective pre-training essential. However, real-world deployments often demand models of variable sizes due to diverse memory and computational constraints, posing…

Computer Vision and Pattern Recognition · Computer Science 2026-03-05 Yucheng Xie , Fu Feng , Ruixiao Shi , Jianlu Shen , Jing Wang , Yong Rui , Xin Geng