A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Filters
Batch Normalization Provably Avoids Rank Collapse for Randomly Initialised Deep Networks
[article]
2020
arXiv
pre-print
Preserved Fulltext
Other Versions
In this work we highlight the fact that batch normalization is an effective strategy to avoid rank collapse for both linear and ReLU networks. ...
Randomly initialized neural networks are known to become harder to train with increasing depth, unless architectural enhancements like residual connections and batch normalization are used. ...
Batch normalization provably prevents rank collapse Since our empirical observations hold equally for both non-linear and linear networks, we here focus on improving the theoretical understanding in the ...
arXiv:2003.01652v3
fatcat:sifsf7zdffahnkkfl7w6bf5uyi
Identifying and Exploiting Structures for Reliable Deep Learning
[article]
2021
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
To do this, we identify structures in deep neural networks that can be exploited to mitigate the above causes of unreliability of deep learning algorithms. ...
The extraordinary performance of these systems often gives the impression that they can be used to revolutionise our lives for the better. ...
Random-GAN is a randomly initialised GAN with the same architecture as SRN-GAN. ...
arXiv:2108.07083v1
fatcat:lducrn5tlfeqvpxevz6gukfvse
Building Compact and Robust Deep Neural Networks with Toeplitz Matrices
[article]
2021
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
Deep neural networks are state-of-the-art in a wide variety of tasks, however, they exhibit important limitations which hinder their use and deployment in real-world applications. ...
When developing and training neural networks, the accuracy should not be the only concern, neural networks must also be cost-effective and reliable. ...
We train our networks for 200 epochs with a batch size of 200. ...
arXiv:2109.00959v1
fatcat:75yhvilerre47leutzanaul67m
Open-Ended Learning Leads to Generally Capable Agents
[article]
2021
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
Other Versions
and Tom Hudson for additional environment art and support. ...
Acknowledgements We would like to thank Simon Osindero, Guy Lever, and Oriol Vinyals for reviewing the manuscript, Satinder Singh and Koray Kavukcuoglu for support of the project, and Marcus Wainwright ...
Racanière et al. (2020) perform a curriculum over environment goals in randomly initialised 2D and 3D worlds. A setter generates goals for a solver agent. ...
arXiv:2107.12808v2
fatcat:wp5lbeezmrb6rdsbqf6etgurtq
Fake BTS Attacks of GSM System on Software Radio Platform
2012
Journal of Networks
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2017; you can also visit the original URL.
The file type is application/pdf.
On the other hand, the keycontrolled puncturing mechanism deletes the parity bits randomly, which ensures a high error correction capability for the Turbo code. ...
The first paper, "Extensive Design for Attack's Recognition and Resistance of Survivable Network" by Hui Xu and Xiang Gu, gives a solution of attack defense of survivable network based on extentics. ...
Unified trust values or a simple trust rank is insufficient for making privacy disclose decision. ...
doi:10.4304/jnw.7.2.275-281
fatcat:3ejjimllcvewjeifopbvhl3zmi
Sample efficiency, transfer learning and interpretability for deep reinforcement learning
2020
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Deep learning has revolutionised artificial intelligence, where the application of increased compute to train neural networks on large datasets has resulted in improvements in real-world applications such ...
Deep reinforcement learning (DRL) has similarly shown impressive results in board and video games, but less so in real-world applications such as robotic control. ...
Classi-
fying Options for Deep Reinforcement Learning. IJCAI Deep Reinforcement Learning
Workshop.
Analysing Deep Reinforcement Learning Agents Trained with Domain Randomisation. ...
doi:10.25560/80978
fatcat:tc4ylokvhzes7muukkvuikmngy
Bayesian Learning for Data-Efficient Control
2017
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
Third, we take a step towards data efficient learning of high-dimensional control using Bayesian neural networks (BNN). ...
From robotics, to finance, to industrial processing, autonomous learning helps obviate a heavy reliance on experts for system identification and controller design. ...
At time t = 0, the path ξ 0 is usually randomly initialised. Otherwise usually MPC initialises ξ t with shifted elements from the previously-optimised path ξ * t−1 . ...
doi:10.17863/cam.16688
fatcat:5gcifx6g3zcs3acdaxpqc4aseu
A complexity evolutionary theory for the emergence of financial and economic crises: Synchronising Gould and Minsky through von Neumann and Mandelbrot
2017
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
CETFEC aims to identify the signals that lead the existence of the necessary conditions for the emergence of crises, rather than trying to predict the timing of crises. ...
before its collapse in 2008. ...
Targeted banks are chosen randomly. ...
doi:10.25560/44548
fatcat:iex5tbgpojf4xmu63xbgoabgba
Crawler
[chapter]
2009
Encyclopedia of Database Systems
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2011; you can also visit the original URL.
The file type is application/pdf.
As a result, it is common (at the time of writing) for data accesses to RAM to require several hundred CPU cycles to resolve. ...
For such workloads, improving the locality of data-intensive operations can have a direct impact on the system's overall performance. ...
and run a batch process on it. ...
doi:10.1007/978-0-387-39940-9_2315
fatcat:x4qspjdytvhvroc7h753dihp7u
Hierarchical bayesian models for sparse signal recovery and sampling
2016
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2018; you can also visit the original URL.
The file type is application/pdf.
This includes extremely low-complexity algorithms for sparse recovery with a powerful analysis framework. ...
Another innovation of this thesis are Bayesian models for signals whose components are known a priori to exhibit a certain statistical trend. ...
Algorithm 9 AMP Algorithm for FFT basis. Initialise: r 0 = y, s 0 = 0. ...
doi:10.25560/32102
fatcat:j534y7z7sjh4pcmtpz5t7ljeoe
On the surprising capacity of linear combinations of embeddings for natural language processing
[article]
2019
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
Acknowledgements Thank particularly to Christof Stocker, the creator of MLDatasets.jl (and numerous other packages), in particular for his bug reports, feature requests and code reviews; and for the initial ...
Acknowledgements We would like to thank Dr Gerhard Wohlgenannt (ITMO University, Saint Petersburg) for his feedback on this work just prior to submission. ...
This, like the word embeddings, is initialised randomly, then trained during the task. ...
doi:10.26182/w0c2-6887
fatcat:hacpoc35tne3pmcfy5wfdl6dsq
Dagstuhl Reports, Volume 3, Issue 09, September 2013, Complete Issue
[article]
2014
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
In the context of the Smart Grid, network overlays can play a key role, for enabling the management of resources in an adaptive fashion. ...
from a Smart Grid metering network. ...
For networks with small highway dimension, one can compute provably small hierarchical labels in polynomial time. ...
doi:10.4230/dagrep.3.9
fatcat:oilweqoffjbx7bjnxvg6vorrzu
Quality of Experience: From Assessment to Application (Dagstuhl Seminar 15022) Understanding Complexity in Multiobjective Optimization (Dagstuhl Seminar 15031) Model-driven Algorithms and Architectures for Self-Aware Computing Systems (Dagstuhl Seminar 15041) Artificial and Computational Intelligence in Games: Integration (Dagstuhl Seminar 15051)
2015
unpublished
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2018; you can also visit the original URL.
The file type is application/pdf.
The example problems mentioned in this report may serve as a first benchmark for such approaches. ...
= 1) to the most important (rank = m); select q criteria randomly with probabilities proportional to the ranks of criteria so that less important objectives have a higher probability of being omitted. ...
The process worked reasonably well, but requires extensive hyper-parameter tuning on the following fronts: (a) neural network weight initialisation, (b) neural network learning rates, (c) exploration rates ...
fatcat:ckaercgotnenhidbwsz7zufttq
An investigation of emergency department overcrowding using data mining and simulation : a patient treatment type perspective
2017
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Scientific Method was selected as an appropriate methodology for the research. ...
This balance is often difficult to achieve and EDs become overwhelmed, resulting in long patient waits, overcrowded treatment areas and excessive stress for ED staff. ...
If the populations are identical in location, the ranks should be randomly mixed between the two samples. ...
doi:10.4225/03/586dc33ad139b
fatcat:odcko7xoj5ckbgplermrsv6t2q
Lisbon, Portugal VALID 2012 Editors VALID 2012 Committee VALID Advisory Chairs VALID 2012 Technical Progam Committee
unpublished
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2018; you can also visit the original URL.
The file type is application/pdf.
We also gratefully thank the members of the VALID 2012 organizing committee for their help in handling the logistics and for their work that is making this professional meeting a success. ...
Despite current solutions, virtualization and abstraction for large scale systems provide less visibility for vulnerability discovery and resolution, and make testing tedious, sometimes unsuccessful, if ...
ACKNOWLEDGMENT The authors would like to thank Krishna Murthy Murlidhar, Sven Neuendorf and Jasmin Zieger for their contributions. ...
fatcat:xg7zeln5evfw3jxncx33bmxyky