Filters








3,623 Hits in 5.8 sec

Training Deep Networks without Learning Rates Through Coin Betting [article]

Francesco Orabona, Tatiana Tommasi
2017 arXiv   pre-print
Instead, we reduce the optimization process to a game of betting on a coin and propose a learning-rate-free optimal algorithm for this scenario.  ...  In this paper, we propose a new stochastic gradient descent procedure for deep networks that does not require any learning rate setting.  ...  Subgradient Descent through Coin Betting In this section, following Orabona and Pal [2016] , we briefly explain how to reduce subgradient descent to the gambling scenario of betting on a coin.  ... 
arXiv:1705.07795v3 fatcat:2tlcskmh3ve4jizoinoz43mz4q

Optimization based Long Short Term Memory Network for Protein Structure Prediction

Pravinkumar Sonsare, Gunavathi C.
2022 U Porto Journal of Engineering  
With the growing attention of deep learning, models such as convolutional neural network and recurrent neural network are also used for this prediction.  ...  In this paper, we proposed a bidirectional embedded recurrent deep neural system using long short term memory (LSTM) cells with continuous coin betting optimizer (COCOB) to tune the hyperparameters for  ...  COCOB is a learning rate free optimizer inspired from a coin betting game.  ... 
doi:10.24840/2183-6493_008.002_0009 doaj:4645ecd5624b43828533403788de7c7e fatcat:esuml6bqcja3tpr23hg3it6vye

Wage against the machine: A generalized deep-learning market test of dataset value

Philip Z. Maymin
2017 International Journal of Forecasting  
It relies on the use of deep learning, comprehensive historical box score statistics, and the existence of betting markets.  ...  How can you tell whether a particular sports dataset really adds value, particularly with regard to betting effectiveness?  ...  The data on betting markets are easily available through a range of sources; the NBA's boxscore and similar data are available through their website; the deep learning algorithm uses the free open-source  ... 
doi:10.1016/j.ijforecast.2017.09.008 fatcat:vtycmgpkj5gl7efqtrnsszg3ju

Deep Learning for Asset Bubbles Detection [article]

Oksana Bashchenko, Alexis Marchal
2020 arXiv   pre-print
We rely on the theory of local martingales in continuous-time and use a deep network to estimate the diffusion coefficient of the price process more accurately than the current estimator, obtaining an  ...  We develop a methodology for detecting asset bubbles using a neural network.  ...  We will do so through the famous example of the doubling strategy. Imagine that a gambler is betting on the outcome of a coin toss. If the coin comes out head, he wins his bet.  ... 
arXiv:2002.06405v1 fatcat:46bww4md2veaxkmnw54wj7myw4

Combining Deep Reinforcement Learning and Search for Imperfect-Information Games [article]

Noam Brown, Anton Bakhtin, Adam Lerer, Qucheng Gong
2020 arXiv   pre-print
The combination of deep reinforcement learning and search at both training and test time is a powerful paradigm that has led to a number of successes in single-agent settings and perfect-information games  ...  This paper presents ReBeL, a general framework for self-play reinforcement learning and search that provably converges to a Nash equilibrium in any two-player zero-sum game.  ...  The value network contains 2 hidden layers with 256 layers each. We train the network with Adam optimizer with learning rate 3 × 10 −4 and halved the learning rate every 400 epochs.  ... 
arXiv:2007.13544v2 fatcat:v7ox7iqki5biloghgfkusezcja

A U-Net Deep Learning Framework for High Performance Vessel Segmentation in Patients With Cerebrovascular Disease

Michelle Livne, Jana Rieger, Orhun Utku Aydin, Abdel Aziz Taha, Ela Marie Akay, Tabea Kossen, Jan Sobesky, John D. Kelleher, Kristian Hildebrand, Dietmar Frey, Vince I. Madai
2019 Frontiers in Neuroscience  
A specialized deep learning method-the U-net-is a promising alternative.  ...  Our work highly encourages the development of clinically applicable segmentation tools based on deep learning.  ...  This is particularly true for the case of deep neural networks with many convolutional layers and many different paths through the network.  ... 
doi:10.3389/fnins.2019.00097 pmid:30872986 pmcid:PMC6403177 fatcat:svpsw6fbunh6zdtgrconcwb37i

Convert index trading to option strategies via LSTM architecture

Jimmy Ming-Tai Wu, Mu-En Wu, Pang-Jen Hung, Mohammad Mehedi Hassan, Giancarlo Fortino
2020 Neural computing & applications (Print)  
The second point is that the trading strategy is difficult to determine the winning rate in the financial market and cannot be brought into the Kelly criterion to calculate the optimal fraction.  ...  Fig. 6 6 An architecture diagram of deep learning neural network Fig. 7 7 An architecture diagram of simple artificial neural network Fig. 8 8 An architecture diagram of deep recurrent neural networks  ...  To further improve the accuracy of prediction, artificial neural networks and deep learning have also been used in financial markets [12, 25, 45] .  ... 
doi:10.1007/s00521-020-05377-6 fatcat:5dfdvfzmb5edvdhdct3luamjzm

AdaS: Adaptive Scheduling of Stochastic Gradients [article]

Mahdi S. Hosseini, Konstantinos N. Plataniotis
2020 arXiv   pre-print
This work attempts to answer a question of interest to both researchers and practitioners, namely "how much knowledge is gained in iterative training of deep neural networks?"  ...  Answering this question introduces two useful metrics derived from the singular values of the low-rank factorization of convolution layers in deep neural networks.  ...  Mapping conditions are shown in Figure 6 and Figure 7 demonstrates the learning rate approximation through AdaS algorithm over successive epoch training.  ... 
arXiv:2006.06587v1 fatcat:6rowxlf73bevdk7zp52423vhx4

Machine Learning and Data mining on the innovation of E-sports industry

2020 International Journal of Education and Information Technologies  
With the help of the agent created by intensive in-depth learning, it can assist players of different levels to carry out routine training, so as to improve the overall activity of the game.  ...  Because the feature detection layer of CNN learns from the training data, the explicit feature extraction is avoided when using CNN, and the neural network can learn from the training data implicitly.  ...  At present, the AI team is mainly based on the deep neural network learning technology of OpenAI five, but compared with the conventional Dota2 game, OpenAI has two main limitations: first, there are 117  ... 
doi:10.46300/9109.2020.14.15 fatcat:ixtomigxafcnbaqmlvojfupppq

XDO: A Double Oracle Algorithm for Extensive-Form Games [article]

Stephen McAleer, John Lanier, Kevin Wang, Pierre Baldi, Roy Fox
2022 arXiv   pre-print
We also introduce Neural XDO (NXDO), where the best response is learned through deep RL.  ...  Policy Space Response Oracles (PSRO) is a reinforcement learning (RL) algorithm for two-player zero-sum games that has been empirically shown to find approximate Nash equilibria in large games.  ...  Both players start with an allotment of coins and each step simultaneously bet an amount of coins. The player who bets the higher amount gets to move the token toward the opponent.  ... 
arXiv:2103.06426v2 fatcat:phkdkzv33vd4johkeybneouoke

Dynamic Filter Networks [article]

Bert De Brabandere, Xu Jia, Tinne Tuytelaars, Luc Van Gool
2016 arXiv   pre-print
By visualizing the learned filters, we illustrate that the network has picked up flow information by only looking at unlabelled training data.  ...  In a traditional convolutional layer, the learned filters stay fixed after training.  ...  [24] learn to rotate a given face to another pose. The authors of [18, 21, 20, 17, 14] train a deep neural network to predict subsequent video frames. Flynn et al.  ... 
arXiv:1605.09673v2 fatcat:5q5ptcahcnadpfkpyrizicfuji

Parabolic Approximation Line Search for DNNs [article]

Maximus Mutschler, Andreas Zell
2021 arXiv   pre-print
A major challenge in current optimization research for deep learning is to automatically find optimal step sizes for each update step.  ...  It surpasses other step size estimating methods and competes with common optimization methods on a large variety of experiments without the need of hand-designed step size schedules.  ...  Funding This research was supported by the German Federal Ministry of Education and Research (BMBF) project 'Training Center Machine Learning, Tübingen' with grant number 01|S17054.  ... 
arXiv:1903.11991v5 fatcat:xqbq747ehvdc5pfq22trme3eme

Gradient-only line searches to automatically determine learning rates for a variety of stochastic training algorithms [article]

Dominic Kafka, Daniel Nicolas Wilke
2020 arXiv   pre-print
to tune the sensitive hyperparameters of learning rate schedules in neural network training.  ...  Gradient-only and probabilistic line searches have recently reintroduced the ability to adaptively determine learning rates in dynamic mini-batch sub-sampled neural network training.  ...  We gratefully acknowledge the NVIDIA corporation, for supporting our research through the NVIDIA GPU grant.  ... 
arXiv:2007.01054v1 fatcat:whphadhorve5xmssblkgsm36x4

A Review on Machine Learning, Artificial Intelligence, and Smart Technology in Water Treatment and Monitoring

Matthew Lowe, Ruwen Qin, Xinwei Mao
2022 Water  
Artificial-intelligence methods and machine-learning models have demonstrated their ability to optimize, model, and automate critical water- and wastewater-treatment applications, natural-systems monitoring  ...  In addition to providing computer-assisted aid to complex issues surrounding water chemistry and physical/biological processes, artificial intelligence and machine-learning (AI/ML) applications are anticipated  ...  As is commonly used with these models, the ANN model relied on backpropagation, meaning that input training data are fed through the model, passing through the output layer, where training error is propagated  ... 
doi:10.3390/w14091384 fatcat:7r5x2rbxmjdkzevzikuvvmoyyq

Blockchain System Defensive Overview for Double-Spend and Selfish Mining Attacks: A Systematic Approach

Kervins Nicolas, Yi Wang, George C. Giakos, Bingyang Wei, Hongda Shen
2020 IEEE Access  
Blockchain is a technology that ensures data security by verifying database of records established in a decentralized and distributed network.  ...  Detection of the selfish mining attack and double-spend attack can be conducted through data analytics using machine learning and deep learning methods.  ...  It generates adversarial examples which are specially crafted noises in the training data sets. They cause deep learning model to make mistakes [137] .  ... 
doi:10.1109/access.2020.3047365 fatcat:6ofb6os2mfea7alfc3l4hnbqly
« Previous Showing results 1 — 15 out of 3,623 results