Filters








1,804 Hits in 4.8 sec

Pretrained AI Models: Performativity, Mobility, and Change [article]

Lav R. Varshney, Nitish Shirish Keskar, Richard Socher
2019 arXiv   pre-print
We then discuss how pretrained models move through actor networks as a kind of computationally immutable mobile, but that users also act as agents of technological change by reinterpreting them via fine-tuning  ...  Further how pretrained models may have a performative effect on society that exacerbates biases.  ...  of how pretrained models move and change through the actions of distinct social groups.  ... 
arXiv:1909.03290v1 fatcat:7doni7tc3rginpokkow2wtiqmy

Robustness of on-device Models: Adversarial Attack to Deep Learning Models on Android Apps [article]

Yujin Huang, Han Hu, Chunyang Chen
2021 arXiv   pre-print
To make it more accessible to end users, many deep learning models are now embedded in mobile apps.  ...  Compared to offloading deep learning from smartphones to the cloud, performing machine learning on-device can help improve latency, connectivity, and power consumption.  ...  Mobile App Security Prior work on AI app security mainly focuses on model protection. Xu et al.  ... 
arXiv:2101.04401v2 fatcat:hknqkmp6njehrifzfbxo5mvm4u

Smart App Attack: Hacking Deep Learning Models in Android Apps [article]

Yujin Huang, Chunyang Chen
2022 arXiv   pre-print
The results call for the awareness and actions of deep learning mobile app developers to secure the on-device models. The code of this work is available at https://github.com/Jinxhy/SmartAppAttack  ...  On-device deep learning is rapidly gaining popularity in mobile applications.  ...  The framework first performs model extraction to a mobile app with a deep learning model and check if it is a fine-tuned model.  ... 
arXiv:2204.11075v1 fatcat:l3l7p6wjsbexnkpylz75m6fboi

Deep Transfer Learning Beyond: Transformer Language Models in Information Systems Research [article]

Ross Gruetzemacher, David Paradice
2021 arXiv   pre-print
Recent progress in natural language processing involving transformer language models (TLMs) offers a potential avenue for AI-driven business and societal transformation that is beyond the scope of what  ...  This is possible because these techniques make it easier to develop very powerful custom systems and their performance is superior to existing methods for a wide range of tasks and applications.  ...  While Fountaine et al. are correct to suggest that organizations need to change their culture to reap the benefits of AI, it is also true that many of the benefits of AI have yet to be realized because  ... 
arXiv:2110.08975v2 fatcat:bw6rzrz2zvdyrgraxoxdraf4d4

Do Better ImageNet Models Transfer Better? [article]

Simon Kornblith, Jonathon Shlens, Quoc V. Le
2019 arXiv   pre-print
An implicit hypothesis in modern computer vision research is that models that perform better on ImageNet necessarily perform better on other vision tasks.  ...  We find that, when networks are used as fixed feature extractors or fine-tuned, there is a strong correlation between ImageNet accuracy and transfer accuracy (r = 0.99 and 0.96, respectively).  ...  Emily Xue for comments on the experiments and manuscript, and Aliza Elkin and members of the Google Brain team for support and ideas.  ... 
arXiv:1805.08974v3 fatcat:fybgvuwa3zhixmbytdeuwdbory

Do Better ImageNet Models Transfer Better?

Simon Kornblith, Jonathon Shlens, Quoc V. Le
2019 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)  
An implicit hypothesis in modern computer vision research is that models that perform better on ImageNet necessarily perform better on other vision tasks.  ...  We find that, when networks are used as fixed feature extractors or fine-tuned, there is a strong correlation between ImageNet accuracy and transfer accuracy (r = 0.99 and 0.96, respectively).  ...  Emily Xue for comments on the experiments and manuscript, and Aliza Elkin and members of the Google Brain team for support and ideas.  ... 
doi:10.1109/cvpr.2019.00277 dblp:conf/cvpr/KornblithSL19 fatcat:nylanmer5zgajd6ry3iyygbu5i

Deep CNN models for predicting COVID-19 in CT and x-ray images

Ahmad Chaddad, Lama Hassan, Christian Desrosiers
2021 Journal of Medical Imaging  
These segmented regions were then employed as an additional input to six deep convolutional neural network (CNN) architectures (AlexNet, DenseNet, GoogleNet, NASNet-Mobile, ResNet18, and DarkNet), pretrained  ...  Incorporating the three ROIs as an additional model inputs further boosts performance to an accuracy of 82.30% and an AUC of 90.10% (DarkNet).  ...  ., AlexNet, GoogleNet, NASNet-Mobile, DenseNet, DarkNet, and ResNet18, pretrained for image classification on the ImageNet dataset.  ... 
doi:10.1117/1.jmi.8.s1.014502 pmid:33912622 pmcid:PMC8071782 fatcat:mtysq47r65by3ftnb3prvpx5vu

Machine Reading Comprehension-Enabled Public Service Information System: A Large-Scale Dataset and Neural Network Models

Changchang Zeng, Shaobo Li, Bin Chen, Mohammad Farukh Hashmi
2022 Wireless Communications and Mobile Computing  
Next, we propose several new neural network models which were continually pretrained on the Chinese public service corpus.  ...  However, they are still far below human performance, indicating that the proposed dataset is challenging.  ...  Human Performance. Table 5 : 5 e evaluation results of pretrained models and human performance.  ... 
doi:10.1155/2022/7088126 fatcat:bfrswwv4trhijphozqhavvxeoi

The great transformer: Examining the role of large language models in the political economy of AI

Dieuwertje Luitse, Wiebke Denkena
2021 Big Data & Society  
In recent years, AI research has become more and more computationally demanding.  ...  This article explores the role LLMs play in the political economy of AI as infrastructural components for AI research and development.  ...  Research into AGI is located on the margins of contemporary AI research which focuses on task-specific applications ('narrow AI').  ... 
doi:10.1177/20539517211047734 fatcat:57lewolflnadhjgynrbz3fmn24

Interoperating Deep Learning models with ONNX.jl

2020 JuliaCon Proceedings  
speed and memory usage in large deep learning models.  ...  The framework makes writing layers as simple as writing mathematical formulae, and it's advanced AD, Zygote [11] , applies automatic differentiation (AD) to calculate derivatives and train the model.  ...  ONNX reduces the friction of moving trained AI models among your favorite tools and frameworks and platforms.  ... 
doi:10.21105/jcon.00059 fatcat:q7ftrjqmxjejdauwvmakdz6ibm

Sub-Character Tokenization for Chinese Pretrained Language Models [article]

Chenglei Si, Zhengyan Zhang, Yingfa Chen, Fanchao Qi, Xiaozhi Wang, Zhiyuan Liu, Yasheng Wang, Qun Liu, Maosong Sun
2021 arXiv   pre-print
Tokenization is fundamental to pretrained language models (PLMs). Existing tokenization methods for Chinese PLMs typically treat each character as an indivisible token.  ...  At the same time, models trained with SubChar tokenizers perform competitively on downstream tasks. We release our code at https://github.com/thunlp/SubCharTokenization to facilitate future work.  ...  We train tokenizers with and without CWS and compare 5.2 Efficiency Improvement the performance of the corresponding pretrained The direct consequence of having more character models.  ... 
arXiv:2106.00400v2 fatcat:rf4u6indwnas3elrc77sqb3hwe

On the Opportunities and Risks of Foundation Models [article]

Rishi Bommasani, Drew A. Hudson, Ehsan Adeli, Russ Altman, Simran Arora, Sydney von Arx, Michael S. Bernstein, Jeannette Bohg, Antoine Bosselut, Emma Brunskill, Erik Brynjolfsson, Shyamal Buch (+102 others)
2021 arXiv   pre-print
AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks.  ...  This report provides a thorough account of the opportunities and risks of foundation models, ranging from their capabilities (e.g., language, vision, robotics, reasoning, human interaction) and technical  ...  Fernando Pereira, Vinodkumar Prabhakaran, Colin Raffel, Marten van Schijndel, Ludwig Schmidt, Yoav Shoham, Madalsa Singh, Megha Srivastava, Jacob Steinhardt, Emma Strubell, Qian Yang, Luke Zettlemoyer, and  ... 
arXiv:2108.07258v2 fatcat:yktkv4diyrgzzfzqlpvaiabc2m

A Roadmap for Big Model [article]

Sha Yuan, Hanyu Zhao, Shuai Zhao, Jiahong Leng, Yangxiao Liang, Xiaozhi Wang, Jifan Yu, Xin Lv, Zhou Shao, Jiaao He, Yankai Lin, Xu Han (+88 others)
2022 arXiv   pre-print
and Application.  ...  In this paper, we cover not only the BM technologies themselves but also the prerequisites for BM training and applications with BMs, dividing the BM review into four parts: Resource, Models, Key Technologies  ...  Still, using a pretraining backbone big vision model improves the performance of segmentation and a larger vision model pretrained with more data typically achieves better performance.  ... 
arXiv:2203.14101v4 fatcat:rdikzudoezak5b36cf6hhne5u4

COVID-19 Modeling: A Review [article]

Longbing Cao, Qing Liu
2021 arXiv   pre-print
The modeling methods involve mathematical and statistical models, domain-driven modeling by epidemiological compartmental models, medical and biomedical analysis, AI and data science in particular shallow  ...  and deep machine learning, simulation modeling, social science methods, and hybrid modeling.  ...  More information about COVID-19 modeling is in https://datasciences.org/covid19-modeling/.  ... 
arXiv:2104.12556v3 fatcat:pj2bketcrveafbjf2m7tx3odxy

A Multichannel Biomedical Named Entity Recognition Model Based on Multitask Learning and Contextualized Word Representations

Hao Wei, Mingyuan Gao, Ai Zhou, Fei Chen, Wen Qu, Yijia Zhang, Mingyu Lu
2020 Wireless Communications and Mobile Computing  
In the previous studies based on deep learning, pretrained word embedding becomes an indispensable part of the neural network models, effectively improving their performance.  ...  The latter achieves the best performance among reported existing feature-based models.  ...  The large performance gains come from taking turns training the target and collaborator submodels. Sachan et al. [28] designed a pretrained BiLSTM model.  ... 
doi:10.1155/2020/8894760 fatcat:z55w6flkovhfnpchlz2lzj2aya
« Previous Showing results 1 — 15 out of 1,804 results