scholarly journals Fine-Tuned Pre-Trained Model for Script Recognition

2021 ◽  
Vol 6 (5) ◽  
pp. 1297-1314
Author(s):  
Mamta Bisht ◽  
Richa Gupta
Keyword(s):  
Alternative Methods ◽  
Text Recognition ◽  
Computational Power ◽  
Preliminary Step ◽  
Script Recognition ◽  
Original Dataset ◽  
Classification Tasks ◽  
Computational Resources ◽  
Numeral Recognition ◽  
Fine Tune

Script recognition is the first necessary preliminary step for text recognition. In the deep learning era, for this task two essential requirements are the availability of a large labeled dataset for training and computational resources to train models. But if we have limitations on these requirements then we need to think of alternative methods. This provides an impetus to explore the field of transfer learning, in which the previously trained model knowledge established in the benchmark dataset can be reused in another smaller dataset for another task, thus saving computational power as it requires to train only less number of parameters from the total parameters in the model. Here we study two pre-trained models and fine-tune them for script classification tasks. Firstly, the VGG-16 pre-trained model is fine-tuned for publically available CVSI-15 and MLe2e datasets for script recognition. Secondly, a well-performed model on Devanagari handwritten characters dataset has been adopted and fine-tuned for the Kaggle Devanagari numeral dataset for numeral recognition. The performance of proposed fine-tune models is related to the nature of the target dataset as similar or dissimilar from the original dataset and it has been analyzed with widely used optimizers.

scholarly journals Optimal Computational Power Allocation in Multi-Access Mobile Edge Computing for Blockchain

Sensors ◽  
2018 ◽  
Vol 18 (10) ◽  
pp. 3472 ◽  
Author(s):  
Yuan Wu ◽  
Xiangxu Chen ◽  
Jiajun Shi ◽  
Kejie Ni ◽  
Liping Qian ◽  
...  
Keyword(s):  
Optimization Problem ◽  
Edge Computing ◽  
Resource Acquisition ◽  
Mobile Edge Computing ◽  
Computational Power ◽  
Mobile Terminals ◽  
Acquisition Costs ◽  
Digital Transactions ◽  
Multi Access ◽  
Computational Resources

Blockchain has emerged as a decentralized and trustable ledger for recording and storing digital transactions. The mining process of Blockchain, however, incurs a heavy computational workload for miners to solve the proof-of-work puzzle (i.e., a series of the hashing computation), which is prohibitive from the perspective of the mobile terminals (MTs). The advanced multi-access mobile edge computing (MEC), which enables the MTs to offload part of the computational workloads (for solving the proof-of-work) to the nearby edge-servers (ESs), provides a promising approach to address this issue. By offloading the computational workloads via multi-access MEC, the MTs can effectively increase their successful probabilities when participating in the mining game and gain the consequent reward (i.e., winning the bitcoin). However, as a compensation to the ESs which provide the computational resources to the MTs, the MTs need to pay the ESs for the corresponding resource-acquisition costs. Thus, to investigate the trade-off between obtaining the computational resources from the ESs (for solving the proof-of-work) and paying for the consequent cost, we formulate an optimization problem in which the MTs determine their acquired computational resources from different ESs, with the objective of maximizing the MTs’ social net-reward in the mining process while keeping the fairness among the MTs. In spite of the non-convexity of the formulated problem, we exploit its layered structure and propose efficient distributed algorithms for the MTs to individually determine their optimal computational resources acquired from different ESs. Numerical results are provided to validate the effectiveness of our proposed algorithms and the performance of our proposed multi-access MEC for Blockchain.

Analyzing self-★ island-based memetic algorithms in heterogeneous unstable environments

2016 ◽  
Vol 32 (5) ◽  
pp. 676-692 ◽  
Author(s):  
Rafael Nogueras ◽  
Carlos Cotta
Keyword(s):  
Memetic Algorithm ◽  
Memetic Algorithms ◽  
Population Based ◽  
Self Healing ◽  
Computational Power ◽  
Ability To Work ◽  
Computing Power ◽  
Simulated Environment ◽  
Computational Resources ◽  
Better Than

Computational environments emerging from the pervasiveness of networked devices offer a plethora of opportunities and challenges. The latter arise from their dynamic, inherently volatile nature that tests the resilience of algorithms running on them. Here we consider the deployment of population-based optimization algorithms on such environments, using the island model of memetic algorithms for this purpose. These memetic algorithms are endowed with self-★ properties that give them the ability to work autonomously in order to optimize their performance and to react to the instability of computational resources. The main focus of this work is analyzing the performance of these memetic algorithms when the underlying computational substrate is not only volatile but also heterogeneous in terms of the computational power of each of its constituent nodes. To this end, we use a simulated environment that allows experimenting with different volatility rates and heterogeneity scenarios (that is, different distributions of computational power among computing nodes), and we study different strategies for distributing the search among nodes. We observe that the addition of self-scaling and self-healing properties makes the memetic algorithm very robust to both system instability and computational heterogeneity. Additionally, a strategy based on distributing single islands on each computational node is shown to perform globally better than placing many such islands on each of them (either proportionally to their computing power or subject to an intermediate compromise).

scholarly journals Is One Teacher Model Enough to Transfer Knowledge to a Student Model?

Algorithms ◽  
2021 ◽  
Vol 14 (11) ◽  
pp. 334
Author(s):  
Nicola Landro ◽  
Ignazio Gallo ◽  
Riccardo La Grassa
Keyword(s):  
Neural Network ◽  
Transfer Learning ◽  
Learning Problem ◽  
Learning Techniques ◽  
Starting Point ◽  
Learning Technique ◽  
Interesting Approach ◽  
Classification Tasks ◽  
Learned Features ◽  
Fine Tune

Nowadays, the transfer learning technique can be successfully applied in the deep learning field through techniques that fine-tune the CNN’s starting point so it may learn over a huge dataset such as ImageNet and continue to learn on a fixed dataset to achieve better performance. In this paper, we designed a transfer learning methodology that combines the learned features of different teachers to a student network in an end-to-end model, improving the performance of the student network in classification tasks over different datasets. In addition to this, we tried to answer the following questions which are in any case directly related to the transfer learning problem addressed here. Is it possible to improve the performance of a small neural network by using the knowledge gained from a more powerful neural network? Can a deep neural network outperform the teacher using transfer learning? Experimental results suggest that neural networks can transfer their learning to student networks using our proposed architecture, designed to bring to light a new interesting approach for transfer learning techniques. Finally, we provide details of the code and the experimental settings.

scholarly journals Uniform Manifold Approximation and Projection Analysis of Soccer Players

Entropy ◽  
2021 ◽  
Vol 23 (7) ◽  
pp. 793
Author(s):  
António M. Lopes ◽  
José A. Tenreiro Machado
Keyword(s):  
Dimensionality Reduction ◽  
Soccer Players ◽  
Intermediate Data ◽  
Professional Soccer ◽  
Computer Visualization ◽  
Original Dataset ◽  
Projection Analysis ◽  
Visualization Tools ◽  
Computational Resources ◽  
Made In

In professional soccer, the choices made in forming a team lineup are crucial for achieving good results. Players are characterized by different skills and their relevance depends on the position that they occupy on the pitch. Experts can recognize similarities between players and their styles, but the procedures adopted are often subjective and prone to misclassification. The automatic recognition of players’ styles based on their diversity of skills can help coaches and technical directors to prepare a team for a competition, to substitute injured players during a season, or to hire players to fill gaps created by teammates that leave. The paper adopts dimensionality reduction, clustering and computer visualization tools to compare soccer players based on a set of attributes. The players are characterized by numerical vectors embedding their particular skills and these objects are then compared by means of suitable distances. The intermediate data is processed to generate meaningful representations of the original dataset according to the (dis)similarities between the objects. The results show that the adoption of dimensionality reduction, clustering and visualization tools for processing complex datasets is a key modeling option with current computational resources.

scholarly journals Adaptive Filter as Efficient Tool for Data Assimilation under Uncertainties

2020 ◽  
Author(s):  
Hong Son Hoang ◽  
Remy Baraille
Keyword(s):  
Kalman Filter ◽  
Dynamical Systems ◽  
Data Assimilation ◽  
Adaptive Filter ◽  
High Dimensional ◽  
Main Principle ◽  
Computational Power ◽  
Observational Errors ◽  
Simulation Results ◽  
Computational Resources

In this contribution, the problem of data assimilation as state estimation for dynamical systems under uncertainties is addressed. This emphasize is put on high-dimensional systems context. Major difficulties in the design of data assimilation algorithms is a concern for computational resources (computational power and memory) and uncertainties (system parameters, statistics of model, and observational errors). The idea of the adaptive filter will be given in detail to see how it is possible to overcome uncertainties as well as to explain the main principle and tools for implementation of the adaptive filter for complex dynamical systems. Simple numerical examples are given to illustrate the principal differences of the AF with the Kalman filter and other methods. The simulation results are presented to compare the performance of the adaptive filter with the Kalman filter.

Surrogate Modeling for Spatially Distributed Fuel Cell Models With Applications to Uncertainty Quantification

2017 ◽  
Vol 14 (1) ◽  
Author(s):  
A. A. Shah
Keyword(s):  
Fuel Cell ◽  
Uncertainty Quantification ◽  
Polymer Electrolyte Membrane ◽  
Cell Voltage ◽  
Surrogate Models ◽  
Alternative Methods ◽  
Spatial Homogeneity ◽  
Electrolyte Membrane ◽  
Spatially Distributed ◽  
Computational Resources

Detailed physics-based computer models of fuel cells can be computationally prohibitive for applications such as optimization and uncertainty quantification. Such applications can require a very high number of runs in order to extract reliable results. Approximate models based on spatial homogeneity or data-driven techniques can serve as surrogates when scalar quantities such as the cell voltage are of interest. When more detailed information is required, e.g., the potential or temperature field, computationally inexpensive surrogate models are difficult to construct. In this paper, we use dimensionality reduction to develop a surrogate model approach for high-fidelity fuel cell codes in cases where the target is a field. A detailed 3D model of a high-temperature polymer electrolyte membrane (PEM) fuel cell is used to test the approach. We develop a framework for using such surrogate models to quantify the uncertainty in a scalar/functional output, using the field output results. We propose a number of alternative methods including a semi-analytical approach requiring only limited computational resources.

scholarly journals A Survey on Efficient Parallelization of Blockchain-based Smart Contracts

2019 ◽  
Vol 3 (5) ◽  
pp. 9-16 ◽  
Author(s):  
Alessio Meneghetti ◽  
Tommaso Parise ◽  
Massimiliano Sala ◽  
Daniele Taufer
Keyword(s):  
State Of The Art ◽  
Promising Method ◽  
Smart Contracts ◽  
Computational Power ◽  
Efficient Management ◽  
Smart Contract ◽  
The Status ◽  
Computational Resources ◽  
Scalable Computation ◽  
Efficient Parallelization

The main problem faced by smart contract platforms is the amount of time and computational power required to reach consensus. In a classical blockchain model, each operation is in fact performed by each node, both to update the status and to validate the results of the calculations performed by others. In this short survey we sketch some state-of-the-art approaches to obtain an efficient and scalable computation of smart contracts. Particular emphasis is given to sharding, a promising method that allows parallelization and therefore a more efficient management of the computational resources of the network.

Permutations and Computational Power: A Molecular Cascade Analysis to Approach Big Data in Psychiatry

2017 ◽  
Vol 41 (S1) ◽  
pp. S55-S55
Author(s):  
A. Drago
Keyword(s):  
False Positive ◽  
Statistical Control ◽  
Computational Power ◽  
Molecular Pathway ◽  
Statistical Knowledge ◽  
Cascade Analysis ◽  
Competing Interest ◽  
Pathway Analyses ◽  
Computational Resources ◽  
Source Of Information

In the last few years, we conducted a number of molecular pathway analyses on the genetic samples provided by the NIMH. The molecular pathway approach accounts for the polygenic nature of the most part of psychiatric disorders. Nevertheless, the limits of this approach including the limited knowledge about the function of the genes, the fact that longer genes have higher probability to harbour variations significantly associated with the phenotype under analysis and the false positive associations for single variations, demand statistical control and bio-statistical knowledge. Permutations are a methodology to control for false positive associations, but their implementation requires that a number of criteria are taken into account: 1) the same number of genes and the same number of variations of the index pathway must be simulated in order to limit the bias of selecting significantly longer or shorter genes; 2) a sufficient number of permutated pathways is created (10E5 to 10E6 depending on computational resources) which demands higher computational power; 3) the correct statistical thresholds are identified and discussed; 4) some pathways might be over-represented and the source of information must be constantly updated. The tools for running a molecular pathway analysis (R Foundation for Statistical Computing, 2013) when interacting with a supercluster PC and the international bioinformatic datasets (Embase, NIMH and others), together with the critical steps of bioinformatics scripting (bash language) are described and discussed.Disclosure of interestThe author has not supplied his declaration of competing interest.

scholarly journals Mask Data Priming Network for Automatic Check-Out

2020 ◽  
Author(s):  
Zhen Xiao ◽  
Jia Zhao ◽  
Gang Sun
Keyword(s):  
Large Scale ◽  
Domain Adaptation ◽  
Ground Truth ◽  
Training Set ◽  
Test Set ◽  
Joint Learning ◽  
Original Dataset ◽  
Annotation Information ◽  
Level Information ◽  
Fine Tune

Auto checkout has received more and more attention in recent years and this system automatically generates a shopping bill by identifying the picture of the products purchased by the customers. However, the system is challenged by the domain adaptation problem, where each image of the training set contains only one commodity, whereas the test set is a collection of multiple commodities. The existing solution to this problem is to resynthesize the training images to enhance the training set. Then the composite images are rendered using CycleGAN to make the image distribution of the training set and the test set more similar. However, we find that the detection boxes given by the ground truth of the common dataset contain a large part of the background area, the area will affect the training process as noise. To solve this problem, we propose a mask data priming method. Specifically, we redo the large scale Retail Product Checkout (RPC) dataset and add segmentation annotation information to each item in the training set image based on the original dataset using pixel-level annotation. Secondly, a new network structure is proposed in which we train the network using joint learning of detectors and counters, and fine-tune the detection network by filtering out suitable images from the test set. Experiments on the RPC dataset have shown that our method yields better results. we used an approach that reached 81.87% compared to 56.68% for the baseline approach which demonstrates that pixel-level information helps to improve the detection results of the network.

scholarly journals Calcium based sorbent calcination and sintering reaction models overview

2018 ◽  
Vol 72 (6) ◽  
pp. 329-339
Author(s):  
Ivan Tomanovic ◽  
Srdjan Belosevic ◽  
Aleksandar Milicevic ◽  
Nenad Crnomarkovic
Keyword(s):  
Efficiency Coefficient ◽  
Computational Power ◽  
The Past ◽  
Reaction Models ◽  
Data Output ◽  
Complex Models ◽  
Internal Particle ◽  
Computational Resources ◽  
Accuracy Data ◽  
Pollutant Gases

Several models considering the pulverized sorbent reactions with pollutant gases were developed over the past years. In this paper, we present a detailed overview of available models for direct furnace injection of pulverized calcium sorbent suitable for potential application in CFD codes, with respect to implementation difficulty and computational resources demand. Depending on the model, variations in result accuracy, data output, and computational power required may occur. Some authors separate the model of calcination reaction, combined with the sintering model, and afterwards model the sulfation. Other authors assume the calcination to be instantaneous, and focus the modelling efforts toward the sulfation reaction, adding the sintering effects as a parameter in the efficiency coefficient. Simple models quantify the reaction effects, while more complex models attempt to describe and explain internal particle reactions through different approaches to modelling of the particle internal structure.

Sign in / Sign up

Export Citation Format

Share Document