Asian Cup Qualification Round 3 Group F stats & predictions

Tecnología e Innovación: Impacto Futuro

A medida que avanza la tecnología, el fútbol también se ve afectado. Exploramos cómo las innovaciones tecnológicas podrían influir en el futuro desarrollo del Grupo F:

JorgeMartinezGlez/thesis<|file_sep|>/manuscript/sections/introduction.tex % !TEX root = ../main.tex section{Introduction} label{sec:introduction} The use of Artificial Intelligence (AI) is becoming more and more common in our society. In particular the use of Machine Learning (ML) algorithms to tackle problems from different domains has become popular. It is possible to find ML algorithms being used in the areas of image recognition cite{krizhevsky2017imagenet}, natural language processing cite{brown2020language}, and even for trading financial instruments cite{wang2019deep}. However the most common area where ML algorithms are being used is in recommender systems cite{he2017survey}. These systems are used by companies such as Netflix cite{netflix} and Amazon cite{amazon} to recommend products or movies to users based on their previous interactions with the system. The implementation of these recommender systems requires that the company has access to large amounts of data about its users' interactions. This access to data is also required when using ML algorithms for research purposes. However access to these datasets is not always easy and there are many restrictions imposed by companies that own them. This problem was already identified in cite{Domingue2017} where the authors discuss the problems that arise from the fact that these datasets are not always available for researchers to use. Even if it is possible for researchers to have access to some of these datasets there is still another problem which is related to the availability of computing resources. These datasets are usually quite large and this makes it hard for researchers to train models on them. For example Amazon's product catalog contains over one billion products cite{amazon}. Training a model on such large datasets requires the use of distributed computing resources which are not always available for researchers. One solution to this problem is the use of cloud computing services. Cloud computing provides users with scalable computing resources which can be used as needed. This allows researchers to train models on large datasets without having to worry about the availability of computing resources. In this thesis we will explore how cloud computing can be used by researchers to train ML models on large datasets. We will also discuss some of the challenges that arise when using cloud computing for this purpose and how these challenges can be addressed. In particular we will focus on the use of cloud computing for training recommender systems on large datasets. subsection{Related Work} label{subsec:related_work} The use of cloud computing for training ML models has been explored in several works. In cite{boyd2018scalable} the authors propose a framework for training deep neural networks on large-scale datasets using cloud computing resources. The framework allows users to specify the number of machines they want to use for training and automatically distributes the data across these machines. In cite{kayalibay2018distributed} the authors propose a distributed training framework for deep learning models that uses cloud computing resources. The framework allows users to specify the number of machines they want to use for training and automatically distributes the data across these machines. Another work that explores the use of cloud computing for training ML models is cite{lee2019deep}. The authors propose a framework that allows users to train deep learning models on large-scale datasets using cloud computing resources. The framework allows users to specify the number of machines they want to use for training and automatically distributes the data across these machines. In addition to these works there are also several works that explore the use of cloud computing for training recommender systems on large-scale datasets. In cite{he2017survey} the authors provide a survey of different approaches for training recommender systems using cloud computing resources. They discuss different techniques such as distributed matrix factorization and distributed deep learning that can be used for this purpose. Another work that explores the use of cloud computing for training recommender systems is cite{kang2018large}. The authors propose a framework that allows users to train recommender systems on large-scale datasets using cloud computing resources. The framework allows users to specify the number of machines they want to use for training and automatically distributes the data across these machines. Finally there are also works that explore other aspects related to using cloud computing for training ML models such as security and privacy issues. In cite{jiang2019secure} the authors propose a secure multi-party computation protocol that allows users to train ML models on private data using cloud computing resources without revealing their data to each other or to the cloud provider. subsection{Thesis Contributions} label{subsec:thesis_contributions} In this thesis we will make several contributions related to using cloud computing for training ML models on large-scale datasets. Firstly we will propose a new framework for training deep learning models on large-scale datasets using cloud computing resources. This framework will build upon existing frameworks such as those proposed in cite{boyd2018scalable,kayalibay2018distributed} but will also include new features such as automatic data preprocessing and model validation. Secondly we will evaluate our proposed framework on several benchmark datasets and compare its performance with existing frameworks. We will also analyze the scalability of our proposed framework by increasing the size of the dataset and measuring its performance. Thirdly we will explore some of the challenges that arise when using cloud computing for training ML models such as security and privacy issues. We will propose solutions to address these challenges such as secure multi-party computation protocols. Finally we will discuss future directions related to using cloud computing for training ML models such as incorporating new machine learning algorithms into our proposed framework.<|file_sepTableofContents<|file_sep>/********************************************************************* * Author: * Jorge Martínez Glez * Alejandro Pina Pina * * Description: * This file contains functions useful when building graphs with dot * *********************************************************************/ #include "utils.h" /* * Returns an edge between two nodes in dot format */ char *edge(char *node1, char *node2) { char *s = malloc(1024); sprintf(s,""%s" -> "%s";n", node1,node2); return s; } /* * Returns an edge between two nodes with attributes in dot format */ char *edge(char *node1,char *node2,char **attrs,int attrs_len) { char **attr = malloc((attrs_len+1)*sizeof(char*)); char *s = malloc(2048); sprintf(s,""%s" -> "%s"",node1,node2); attr[0] = s; int i; char buf[128]; sprintf(buf,"["); strcat(s,buf); int len=1; int first=true; int j; int k; int count=0; // Add all attributes for(i=0;i=128)