10 Hits in 7.8 sec

Compressing Language Models using Doped Kronecker Products [article]

Urmish Thakker, Paul N. Whatmough, Zhi-Gang Liu, Matthew Mattina, Jesse Beu
2020 arXiv   pre-print
More formally, we propose doping, a process of adding an extremely sparse overlay matrix on top of the pre-defined KP structure. We call this compression method doped kronecker product compression.  ...  We present experimental results that demonstrate compression of a large language model with LSTM layers of size 25 MB by 25x with 1.4% loss in perplexity score.  ...  Similar observation can be made for Solution 2 proposed in this paper -Doping Add a sparse matrix (94% sparsity) to A to provide additional degrees of freedom to the elements in matrix A.  ... 
arXiv:2001.08896v5 fatcat:pwyear3j3vepphqrh6zwp7eapm

An Improved Method for Extractive Based Opinion Summarization Using Opinion Mining

Surbhi Bhatia, Mohammed AlOjail
2022 Computer systems science and engineering  
Short-Term Memory (RNN-LSTM) deep learning model for short and exact synopsis using seq2seq model.  ...  It presents a paradigm shift with regard to the way extractive summaries are generated. Novel algorithms for word extraction using assertions are proposed.  ...  They have used Rouge tool for fusion of sentences to incorporate the rules of NLP. Their work is limited to fusion of sentences.  ... 
doi:10.32604/csse.2022.022579 fatcat:3xdyvgkrzjfx5hv7v33d7braum

Auto-Differentiating Linear Algebra [article]

Matthias Seeger, Asmus Hetzel, Zhenwen Dai, Eric Meissner, Neil D. Lawrence
2019 arXiv   pre-print
Development systems for deep learning (DL), such as Theano, Torch, TensorFlow, or MXNet, are easy-to-use tools for creating complex neural network models.  ...  We detail how a number of matrix decompositions (Cholesky, LQ, symmetric eigen) can be implemented as differentiable operators.  ...  a full-fledged Kalman filtering implementation, using our novel operators.  ... 
arXiv:1710.08717v5 fatcat:hcdzpjjowvfo5btejq3w452afy

Machine Learning Applications for Short Reach Optical Communication

Yapeng Xie, Yitong Wang, Sithamparanathan Kandeepan, Ke Wang
2022 Photonics  
In particular, taking advantage of their high accuracy, adaptability, and implementation efficiency, ML has been widely studied in short-reach optical communications for optical performance monitoring  ...  With the rapid development of optical communication systems, more advanced techniques conventionally used in long-haul transmissions have gradually entered systems covering shorter distances below 100  ...  The basic structure of LSTM is shown in Figure 6.  ... 
doi:10.3390/photonics9010030 fatcat:n436p2z4vvbhlltafonxxnai5y

Towards the Digitalisation of Porous Energy Materials: Evolution of Digital Approaches for Microstructural Design

Zhiqiang Niu, Valerie Pinfield, Billy Wu, Huizhi Wang, Kui Jiao, Dennis Y. C. Leung, Jin Xuan
2021 Energy & Environmental Science  
Porous energy materials are essential components of many energy devices and systems, the development of which have been long plagued by two main challenges. The first is the 'curse of...  ...  Acknowledgements The authors would like to acknowledge the support of the Royal Society -K. C.  ...  China (51861130359) and the UK Royal Society (NAF\R1\180146), the Faraday Institution Multi-Scale Modelling project ((EP/S003053/1, grant number FIRG003), and the support from Engineering and Physical  ... 
doi:10.1039/d1ee00398d fatcat:tqqvy4vbs5gkdfn4x6frrosxuu

Machine learning for intelligent optical networks: A comprehensive survey

Rentao Gu, Zeyuan Yang, Yuefeng Ji
2020 Journal of Network and Computer Applications  
In this paper, a detailed survey of existing applications of ML for intelligent optical networks is presented.  ...  ML techniques are proved to have superiority on solving complex problems; and thus recently, ML techniques have been used for many optical network applications.  ...  Related Enabling Techniques In addition to the inherent advantages of ML techniques, there are existed enabling techniques, from aspects of network architecture, network management, and optical devices  ... 
doi:10.1016/j.jnca.2020.102576 fatcat:yxrh4phho5d5hggqwwh7sino3i


2020 2020 IEEE International Conference on Consumer Electronics - Taiwan (ICCE-Taiwan)  
Identifying the type of a camera used to capture an investigated image is a useful image forensic tool, which usually employs machine learning or deep learning techniques to train the source camera models  ...  LSTM which can maintain a long-term memory is effective for training such features. Introduction of LSTM into important scene detection of baseball videos is the biggest contribution of this paper.  ...  Our model uses a layer of RNN to provide the function of identifying images, and a layer of LSTM to analyze and generate subtitles.  ... 
doi:10.1109/icce-taiwan49838.2020.9258230 fatcat:g25vw7mzvradxna2grlzp6kgiq


2021 2021 IEEE International Conference on Consumer Electronics-Taiwan (ICCE-TW)  
using the compressed sensing technique.  ...  As the quantization technique greatly reduces the data bit-width, the indexing cost may account for a large proportion of data traffic after sparse compression.  ... 
doi:10.1109/icce-tw52618.2021.9602919 fatcat:aetmvxb7hfah7iuucbamos2wgu


2021 Basic & Clinical Pharmacology & Toxicology  
Conclusion: Emergency nursing model can play a significant role in emergency of critically ill patients.  ...  Objective: The study was to investigate the application of emergency nursing model in emergency treatment in critically ill patients.  ...  Objective: A thulium doped fiber laser is designed, and a narrow linewidth laser is obtained by using the laser linewidth compression technology.  ... 
doi:10.1111/bcpt.13588 pmid:34041856 fatcat:wobg4mjgunfilgvzqsj5l4an7m

Table of contents

2019 2019 Chinese Control Conference (CCC)   unpublished
SU Yin, YANG Cuili, QIAO Junfei 1598 Modeling of Hysteresis for Piezo-driven Stages Using a Rate-dependent Krasnosel'skii-Pokrovskii Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  ...  LV Wang, HU Songlin, CHEN Xiaoli, CHENG Zihao, GE Hui 5368 Guaranteed Structural Controllability for Networked Systems with Minimum Input/Edge Addition . . . . . . . . . . . . . . . . . . . . . . . . .  ... 
doi:10.23919/chicc.2019.8866607 fatcat:bkwxxogrqvcmfpmqiek3xgnuoe