no code implementations • 6 Nov 2017 • Xiaoliang Dai, Hongxu Yin, Niraj K. Jha
To address these problems, we introduce a network growth algorithm that complements network pruning to learn both weights and compact DNN architectures during training.
no code implementations • 30 May 2018 • Xiaoliang Dai, Hongxu Yin, Niraj K. Jha
To address these problems, we propose a hidden-layer LSTM (H-LSTM) that adds hidden layers to LSTM's original one level non-linear control gates.
1 code implementation • CVPR 2019 • Xiaoliang Dai, Peizhao Zhang, Bichen Wu, Hongxu Yin, Fei Sun, Yanghan Wang, Marat Dukhan, Yunqing Hu, Yiming Wu, Yangqing Jia, Peter Vajda, Matt Uyttendaele, Niraj K. Jha
We formulate platform-aware NN architecture search in an optimization framework and propose a novel algorithm to search for optimal architectures aided by efficient accuracy and resource (latency and/or energy) predictors.
no code implementations • 30 Jan 2019 • Hongxu Yin, Guoyang Chen, Yingmin Li, Shuai Che, Weifeng Zhang, Niraj K. Jha
In this work, we propose a hardware-guided symbiotic training methodology for compact, accurate, yet execution-efficient inference models.
no code implementations • 27 May 2019 • Xiaoliang Dai, Hongxu Yin, Niraj K. Jha
Deep neural networks (DNNs) have become a widely deployed model for numerous machine learning applications.
no code implementations • 11 Oct 2019 • Hongxu Yin, Bilal Mukadam, Xiaoliang Dai, Niraj K. Jha
For server (edge) side inference, we achieve a 96. 3% (95. 3%) accuracy in classifying diabetics against healthy individuals, and a 95. 7% (94. 6%) accuracy in distinguishing among type-1/type-2 diabetic, and healthy individuals.
2 code implementations • CVPR 2020 • Hongxu Yin, Pavlo Molchanov, Zhizhong Li, Jose M. Alvarez, Arun Mallya, Derek Hoiem, Niraj K. Jha, Jan Kautz
We introduce DeepInversion, a new method for synthesizing images from the image distribution used to train a deep neural network.
no code implementations • 18 Apr 2020 • Wenhan Xia, Hongxu Yin, Niraj K. Jha
These large, deep models are often unsuitable for real-world applications, due to their massive computational cost, high memory bandwidth, and long latency.
no code implementations • 29 Jul 2020 • Wenhan Xia, Hongxu Yin, Xiaoliang Dai, Niraj K. Jha
Modern deep neural networks are powerful and widely applicable models that extract task-relevant information through multi-level abstraction.
no code implementations • 20 Feb 2021 • Shayan Hassantabar, Joe Zhang, Hongxu Yin, Niraj K. Jha
At the patient level, MHDeep DNNs achieve an accuracy of 100%, 100%, and 90. 0% for the three mental health disorders, respectively.
2 code implementations • CVPR 2021 • Hongxu Yin, Arun Mallya, Arash Vahdat, Jose M. Alvarez, Jan Kautz, Pavlo Molchanov
In this work, we introduce GradInversion, using which input images from a larger batch (8 - 48 images) can also be recovered for large networks such as ResNets (50 layers), on complex datasets such as ImageNet (1000 classes, 224x224 px).
no code implementations • CVPR 2021 • Yerlan Idelbayev, Pavlo Molchanov, Maying Shen, Hongxu Yin, Miguel A. Carreira-Perpinan, Jose M. Alvarez
We study the problem of quantizing N sorted, scalar datapoints with a fixed codebook containing K entries that are allowed to be rescaled.
no code implementations • 12 Jul 2021 • Pavlo Molchanov, Jimmy Hall, Hongxu Yin, Jan Kautz, Nicolo Fusi, Arash Vahdat
We analyze three popular network architectures: EfficientNetV1, EfficientNetV2 and ResNeST, and achieve accuracy improvement for all models (up to $3. 0\%$) when compressing larger models to the latency level of smaller models.
no code implementations • 13 Jul 2021 • Xin Dong, Hongxu Yin, Jose M. Alvarez, Jan Kautz, Pavlo Molchanov, H. T. Kung
Prior works usually assume that SC offers privacy benefits as only intermediate features, instead of private data, are shared from devices to the cloud.
no code implementations • 29 Sep 2021 • Pavlo Molchanov, Jimmy Hall, Hongxu Yin, Jan Kautz, Nicolo Fusi, Arash Vahdat
In the second phase, it solves the combinatorial selection of efficient operations using a novel constrained integer linear optimization approach.
1 code implementation • CVPR 2023 • Huanrui Yang, Hongxu Yin, Maying Shen, Pavlo Molchanov, Hai Li, Jan Kautz
This work aims on challenging the common design philosophy of the Vision Transformer (ViT) model with uniform dimension across all the stacked blocks in a model stage, where we redistribute the parameters both across transformer blocks and between different structures within the block via the first systematic attempt on global structural pruning.
1 code implementation • 20 Oct 2021 • Maying Shen, Hongxu Yin, Pavlo Molchanov, Lei Mao, Jianna Liu, Jose M. Alvarez
We propose Hardware-Aware Latency Pruning (HALP) that formulates structural pruning as a global resource allocation optimization problem, aiming at maximizing the accuracy while constraining latency under a predefined budget.
no code implementations • CVPR 2022 • Maying Shen, Pavlo Molchanov, Hongxu Yin, Jose M. Alvarez
Through extensive experiments on ImageNet, we show that EPI empowers a quick tracking of early training epochs suitable for pruning, offering same efficacy as an otherwise ``oracle'' grid-search that scans through epochs and requires orders of magnitude more compute.
1 code implementation • CVPR 2022 • Hongxu Yin, Arash Vahdat, Jose Alvarez, Arun Mallya, Jan Kautz, Pavlo Molchanov
A-ViT achieves this by automatically reducing the number of tokens in vision transformers that are processed in the network as inference proceeds.
Ranked #34 on Efficient ViTs on ImageNet-1K (with DeiT-S)
no code implementations • 14 Feb 2022 • Ali Hatamizadeh, Hongxu Yin, Pavlo Molchanov, Andriy Myronenko, Wenqi Li, Prerna Dogra, Andrew Feng, Mona G. Flores, Jan Kautz, Daguang Xu, Holger R. Roth
Federated learning (FL) allows the collaborative training of AI models without needing to share raw data.
no code implementations • CVPR 2022 • Ali Hatamizadeh, Hongxu Yin, Holger Roth, Wenqi Li, Jan Kautz, Daguang Xu, Pavlo Molchanov
In this work we demonstrate the vulnerability of vision transformers (ViTs) to gradient-based inversion attacks.
8 code implementations • 20 Jun 2022 • Ali Hatamizadeh, Hongxu Yin, Greg Heinrich, Jan Kautz, Pavlo Molchanov
Pre-trained GC ViT backbones in downstream tasks of object detection, instance segmentation, and semantic segmentation using MS COCO and ADE20K datasets outperform prior work consistently.
Ranked #132 on Semantic Segmentation on ADE20K
1 code implementation • 13 Oct 2022 • Maying Shen, Hongxu Yin, Pavlo Molchanov, Lei Mao, Jianna Liu, Jose M. Alvarez
We propose Hardware-Aware Latency Pruning (HALP) that formulates structural pruning as a global resource allocation optimization problem, aiming at maximizing the accuracy while constraining latency under a predefined budget on targeting device.
1 code implementation • CVPR 2023 • Paul Micaelli, Arash Vahdat, Hongxu Yin, Jan Kautz, Pavlo Molchanov
Our Landmark DEQ (LDEQ) achieves state-of-the-art performance on the challenging WFLW facial landmark dataset, reaching $3. 92$ NME with fewer parameters and a training memory cost of $\mathcal{O}(1)$ in the number of recurrent modules.
Ranked #2 on Face Alignment on WFLW
2 code implementations • 9 Jun 2023 • Ali Hatamizadeh, Greg Heinrich, Hongxu Yin, Andrew Tao, Jose M. Alvarez, Jan Kautz, Pavlo Molchanov
At a high level, global self-attentions enable the efficient cross-window communication at lower costs.
no code implementations • CVPR 2023 • Divyam Madaan, Hongxu Yin, Wonmin Byeon, Jan Kautz, Pavlo Molchanov
We propose a novel framework and a solution to tackle the continual learning (CL) problem with changing network architectures.
no code implementations • 25 Jun 2023 • Anna Bair, Hongxu Yin, Maying Shen, Pavlo Molchanov, Jose Alvarez
Robustness and compactness are two essential attributes of deep learning models that are deployed in the real world.
no code implementations • 29 Aug 2023 • Yazhou Xing, Amrita Mazumdar, Anjul Patney, Chao Liu, Hongxu Yin, Qifeng Chen, Jan Kautz, Iuri Frosio
We present a learning-based system to reduce these artifacts without resorting to complex acquisition mechanisms like alternating exposures or costly processing that are typical of high dynamic range (HDR) imaging.
no code implementations • 2 Oct 2023 • Jingwei Sun, Ziyue Xu, Hongxu Yin, Dong Yang, Daguang Xu, Yiran Chen, Holger R. Roth
However, applying FL to finetune PLMs is hampered by challenges, including restricted model parameter access, high computational requirements, and communication overheads.
2 code implementations • 12 Dec 2023 • Ji Lin, Hongxu Yin, Wei Ping, Yao Lu, Pavlo Molchanov, Andrew Tao, Huizi Mao, Jan Kautz, Mohammad Shoeybi, Song Han
Visual language models (VLMs) rapidly progressed with the recent success of large language models.
Ranked #23 on Visual Question Answering on MM-Vet
4 code implementations • 14 Feb 2024 • Shih-Yang Liu, Chien-Yi Wang, Hongxu Yin, Pavlo Molchanov, Yu-Chiang Frank Wang, Kwang-Ting Cheng, Min-Hung Chen
By employing DoRA, we enhance both the learning capacity and training stability of LoRA while avoiding any additional inference overhead.
no code implementations • 4 Mar 2024 • Qiushan Guo, Shalini De Mello, Hongxu Yin, Wonmin Byeon, Ka Chun Cheung, Yizhou Yu, Ping Luo, Sifei Liu
Vision language models (VLMs) have experienced rapid advancements through the integration of large language models (LLMs) with image-text pairs, yet they struggle with detailed regional visual understanding due to limited spatial awareness of the vision encoder, and the use of coarse-grained training data that lacks detailed, region-specific captions.
1 code implementation • 27 Mar 2024 • De-An Huang, Shijia Liao, Subhashree Radhakrishnan, Hongxu Yin, Pavlo Molchanov, Zhiding Yu, Jan Kautz
In addition to leveraging existing video datasets with timestamps, we propose a new task, Reasoning Temporal Localization (RTL), along with the dataset, ActivityNet-RTL, for learning and evaluating this task.