Integration of Neonatal Mortality Prediction Models into a Clinical Decision Support System

Integration of Neonatal Mortality Prediction Models into a Clinical Decision Support System
by
Hasmik Martirosyan, B.Sc.

A thesis submitted to the Faculty of Graduate and Postdoctoral Affairs
in partial fulfillment of the requirements for the degree of
Master of Applied Science
in
Biomedical Engineering
Ottawa - Carleton Institute for Biomedical Engineering (OCIBME)
Carleton University
Ottawa, Ontario © 2015

https://curve.carleton.ca/system/files/theses/32040.pdf

Yenidoğan Ölüm Tahmin Modellerinin Klinik Karar Destek Sistemine  Entegrasyonu.

Abstract
This thesis describes the development of neonatal mortality risk estimation models using Artificial Neural Networks (ANNs), the integration of these models into the Physician-Parent Decision Support (PPADS) tool, and the pilot study to test the PPADS tool.
A set of data mining programs were created to automate the data preparation, the development of ANN models and the selection of models that satisfy the usefulness criteria specified by our clinician partners. These programs were used to classify neonatal mortality data (6% mortality rate) with the average sensitivity and specificity of 81% and 98% respectively.
The mortality models were integrated with the PPADS tool to provide predictions about the risk of mortality for neonates admitted to the Neonatal Intensive Care Unit (NICU).
The observational and survey study conducted with parents whose infant did not graduate (died) from the NICU gave encouraging results regarding the usefulness of the PPADS tool.
....

Fast Artificial Neural Network library
In order to create ANN models, and make predictions in real time environment, we have explored several open sources libraries. We found the Fast Artificial Neural Network (FANN) library to be suitable to our work for the following reasons: firstly, the library implements feed-forward networks, which our research group has identified to be well performing machine learning methods for our medical datasets; secondly, the fast performance is one of the main features of the library, which is important in processing and analyzing real-time data; lastly, the library is implemented in C language which makes the FANN library based applications compatible with many software environments and portable to many different computer architectures or platforms (Nissen, 2007), (FANN, 2014).
....

Read More

Artificial Neural Network Approach to Mobile Robot Localization

Artificial Neural Network Approach to Mobile Robot Localization
David Swords
Aalto University / School of Electrical Engineering
Department of Automation and Systems Technology
Thesis submitted in partial fulfillment of the requirements for the degree of
Master of Science in Technology

In the robotics community, localization is considered a solved problem, however, the topic is still open to investigation. Mobile robot localization has been focused on developing low-cost approaches and there has been great success using probabilistic methods. Parallel to this, and to a much lesser extent, artificial neural networks (ANNs) have been applied to the problem area with varying success.

A system is proposed in this thesis where the typical probabilistic approach is replaced with one based purely on ANNs. This type of localization attempts to harness the simplicity, scalability and adaptability that ANNs are known for. The ANN approach allows for the encapsulation of a number of steps and elements well known in a probabilistic approach, resulting in the elimination of an internal explicit map, providing pose estimate on network output and network update at runtime.

First, a coordinate-based approach to localization is explored: 1D and 2D trained maps with pose estimates. Second, the coordinate-based approach is eliminated in an effort to replicate a more biologically inspired localization. Finally, a path-finding algorithm applying the new localizaiton approach is presented
....
The FANN library (FANN) provides tools implementing, designing, visualizing
and simulating ANNs.
...
FANN provides many commandline functions and visualization tools for creating,
training and simulating ANNs. With the availablity of visualization tools,
like FANNTool, make it much easier to deveop ANNs and applying them to
common tasks like curve fitting, pattern recognition, and clustering.
...
FANNTool is a GUI to the FANN library which allows ease of use without the
need to program the function calls necessary to perform simple tasks. FANNTool
is an open source project, supported by a community of FANN users.

Read More

ESTIMATIVE OF ULTRAVIOLET SOLAR RADIATION IN BOTUCATU/SP/BRAZIL UTILIZING MACHINE LEARNING TECNICS.

Botucatu, 2013. XXXp. Dissertação (Mestrado em Agronomia / Energia na Agricultura) – Faculdade de Ciências Agronômicas, Universidade Estadual Paulista.
Author: THIAGO DO NASCIMENTO SANTANA DE ALMEIDA

SUMMARY
In this papper is evaluated the estimation of daily solar ultraviolet radiation (UV) using machine learning techniques in Botucatu / SP / Brazil. To develop the model was utilized the artificial neural networks with linear function, the support vector machine with linear function and with RBF function. As input to each of the techniques, were tested five groups containing different weather variables measured as routine in Botucatu radiometry solar station.

Read More

Rapport de Projet VidéoBuzz

 Rapport de Projet Master -Université d'Avignon

Description

Les sites d'hébergement de vidéos déposées par les utilisateurs ont connu un très fort développement ces dernières années. Certaines de ces vidéos ont été consulté par des milliers d'internautes, d'autres restent archivées sans qu'elles soient significativement visionnées. Ce projet consiste à développer un système de prédiction de la fréquentation à court terme d'une vidéo. Cette prédiction pourra utiliser 2 indices complémentaires :
- à partir d'un historique de la fréquentation d'une sur une période donnée (par exemple de 2 jours), on pourra utiliser un système prédictif (par exemple un réseau de neurones) pour faire une estimation de la fréquentation future,
- les utilisateurs ajoutent souvent des méta-données (des tags) aux vidéos déposées. Ces informations et le titre de la vidéo peuvent donner une indication sur le buzz potentiellement généré par une vidéo.

Outre les propositions de méthodes et leur évaluation objectives, le projet consiste à développer une application Web permettant de visualiser les prédictions réalisées.

Fanntool ; 

 

FannTool est un logiciel multi-plateforme développé en C qui fournit un ensemble d’outils permettant l’utilisation de la librairie FANN. FANN ( pour Fast Artificial Neural Network ) est une librairie open-source de réseaux de neurones qui permet à la fois l'entraînement du réseau neuronal et ensuite l’utilisation de celui-ci.
Il faut lui fournir en entrer un fichier respectant ce format:

Read More

Efficient Segmentation and Classification of Hyper Spectral Cubes

Efficient Segmentation and Classification of Hyper Spectral Cubes

Papagrigoriou Stylianos
Technical University of Crete

Abstract:

Hyper Spectral Imaging is a powerful analytical tool, which has been used in a wide area of applications, from Satellite Imaging to Biomedical Diagnosis. Spectral imagery of either macroscopic or microscopic origin is usually depicted in a spectral cube, a registered set of images, featuring one spectral and two spatial dimensions as pixel coordinates. From each pixel, associated with a spectrum -instead of an RGB value-, one is able to extract information about the nature of the material, by studying its spectral signature on the Spectral Cube.
This technique offers a non-destructive and non-invasive way (one does not have to extract part of the material and bring it to the lab) of examining materials, suitable for medical purposes. In the hereby thesis the computational capabilities of spectral imaging methods are examined and attempted to be improved, in order to provide real time pixel classification. Specifically, a successful attempt is made to create a hyper spectral classifier with real-time performance for cubes acquired from a cervix biopsy. Various techniques are tested for efficient segmentation of the Cube, in order to generate a golden standard for the training process. The classification is performed using Neural Networks while the final result is a GPU implementation, the main reason behind the speed up of the application.
Although this study was based on specific medical data, it is possible to be generalized on any aspect of Hyper Spectral Imaging, and shows that real-time Hyper Spectral Processing for classification purposes is feasible.

FannTool ;

...
FANN library, provides a really fast and powerful tool -FannTool- for training neural networks with many different algorithms and automatic mechanisms for identifying the best activation function and training algorithm. FANN library was used exactly for these two purposes (finding the best activation function and training algorithm) and train the Network
...

Read More

Filyos Hisarönü Dalga Verilerinin Yayap Sinir Ağları, Arima Modelleri ve Melez Modeller ile Tahmini

Gazi Üniversitesi
Fen Bilimleri Enstitüsü
İnşaat Mühendisliği
Yüksek Lisans Tezi
MART 2010
ANKARA
Turan GÜZEL

Filyos Hisarönü Dalga Verilerinin Yayap Sinir Ağları, Arima Modelleri ve Melez  Modeller ile Tahmini

Özet :

Zaman serilerinin çözümünde birçok yaklaşım kullanılmaktadır. Kıyı yapılarının projelendirilmesinde, özellikle proje sahasında yapılan ölçümlerden elde edilen verilerin kullanılması yapılacak projenin devamlılığı açısından önemlidir. Bu çalışmada planlanan Filyos Limanı için yörede yapılan dalga yüksekliği ölçüm verileri  kullanılmıştır. Bu verilerin oluşturduğu zaman serisi elde olmayan sebeplerden dolayı kesintilere uğramıştır. Eksik kalan dalga yüksekliği verileri Yapay Sinir Ağları, ARIMA Modelleri ve Melez Modeller kullanılarak tamamlanmıştır. Yapay Sinir Ağı Modeli seçilen zaman serisine en iyi uyumu sağlamasına rağmen, ARIMA Modeli diğer serilere Yapay Sinir Ağı modelinden daha yakın sonuçlar verdiğinden dolayı göz ardı edilmemelidir.

Abstract:

Many approaches are used in the solution of time series. Especially measurements of projects field of the data use during projecting coastal structures. This is important for term of continuity of structures. In this study, wave height data are taken in Filyos Harbor field. These wave heights are created time series. This time series are disrupted by unaccountable causes. Missing wave heights are completed by using Artificial Neural Network Model, ARIMA Model and Hybrid Model. Artificial Neural Network Model is best fit to provided selected time series. Although ARIMA Model is more provide results then Artificial Neural Network Model to the other series.

FannTool;

...
Çalışmada, STATGRAPHICS Centurion programı ARIMA modellemeleri için
ve FannTool programı yapay sinir ağı modellemeleri için kullanılmıştır.
...

Read More

A Back Propagation Artificial Neural Networks Approach to Automatic Customer Record Classification

Maastricht University / Department of Quantitative Economics
Master Thesis
A Back Propagation Artificial Neural Networks Approach to Automatic Customer Record Classification.
Author: Camilo Gaviria

Introduction :

In this thesis I address the problem of finding the correct customer to account (AMID)
assignment based on several variables called customer attributes. These customer at-
tributes contain information that provide a partial description of the customer, i.e. cus-
tomer name, customer address, etc. As each customer record carries many attributes,
and there are many possible accounts or classes to which it can belong, this is in itself
a combinatorial optimization problem which is hard to solve. I will call the problem
of minimizing the number of incorrect class (AMID) assignments from now on, the
customer alignment problem.
...

In the literature, many approach to address similar problems are discussed,
one of these, which I have chosen to explore, is the utilization of neural network based
classi cation approaches that are commonly used in the classi cation of natural language
documents as patents, articles, books. In this thesis I will approach the customer record
alignment problem using an adaptation of the methodology used by to classify
patent documents.

FannTool ;

...
Another advantage of using this pre-compiled library, is the availability of a series of
open source graphical user interfaces. By using one of these GUI's, the development
efort is significantly reduced, and the number of additional tests and configurations
that can be easily performed is much larger. Taking into account that there are several
parameters for which there is no one setup rule (number of neurons per layer, number of layer, type of training algorithm, etc.) being able to easily change the configuration
of the ANN is an highly valued feature. The GUI of choice was FANNTool 1.3 an
open source GUI developed specifically for FANN and portable to several OS. This GUI
allows for direct instance scaling, data manipulation as well as real time monitoring of
the learning process.

...

Read More

Yapay Sinir Ağları Metodolojisi ile Düzeltme Yöntemi Öngörüsü Yapısal İş İstatistikleri Anketi İçin Bir Uygulama

Yapay Sinir Ağları Metodolojisi ile Düzeltme Yöntemi Öngörüsü Yapısal İş İstatistikleri Anketi İçin Bir Uygulama
Ahmet Mert AKTAS
Türkiye İstatistik Kurumu
Uzmanlık Tezi - 2010

Özet:

Günümüzde, teknolojinin gelismesi ile bilgisayarlar karar verici organlar olmaya
baslamıs ve insan zekasına haiz birçok durumu bir arada değerlendirme yeteneğine,
çesitli algoritmalar ile erismislerdir. İnsan zekası benzeri bir yapıyı, bilgisayara aktarma
konusunda yazılım yöntemlerinden biri de yapay sinir ağları ve yapay zeka’dır.
Bu çalısmada, yapay sinir ağlarının düzeltme yöntemi olarak kullanılmasıyla ilgili
kapsamlı bir arastırma olusturulması amaçlanmıstır. Öncelikle yapay zeka kavramından
bahsedilmis, yapay zekanın bir alt baslığı olan yapay sinir ağları kullanılarak Yapısal İş
İstatistikleri Ana Soru Kağıdı (A101) için düzeltme yöntemi öngörülerinde
bulunulmustur. Yapmıs olduğumuz bu öngörülerin somut verilerle elde edilen basarısı
değerlendirilmistir.
Özellikle soru kağıdı üzerindeki nitelikli isgücünü arttırması ve cevaplayıcı yükünü
azaltması gibi belirgin avantajları nedeniyle yapay sinir ağlarını, soru kağıtları için
düzeltme yöntemi olarak kullanmanın gerekliliği çok önemlidir. Standart uygulamaların
altyapısı ancak insani unsurların tam kontrollü sistemlere aktarılmasıyla mümkün
olacaktır. Ancak yıllık olarak derlenen soru kağıtlarından elde edilen verilerin su an için
sınırlı sayıda olması ve ilgili periyotta birçok farklı değiskenden etkilenmesi gibi
nedenlerle uygulama üzerinde daha çok arastırma yapılmalıdır.

FannTool Hakkında;

...
FannTool, hem menüler yardımıyla hem de menü içeriği ile bire bir tanımlanmıs komut
düğmeleriyle kullanılabilmektedir. FannTool program içeriği ve yapılandırması daha
genis kitlelere ulasması için İngilizce olarak hazırlanmıstır. Program yardımıyla
istenilen eğitim metotları ile eğitim gerçeklestirilebileceği gibi program optimum eğitim
algoritmasını da uygulama için tespit edebilmektedir. Benzer sekilde gizli katman ve
çıkıs katmanları için en yaygın fonksiyonlar programda yer almaktadır. Bu fonksiyonlar
arasından seçim yapılabileceği gibi, program aktivasyon fonksiyonları içinde optimum
yöntemi tespit eden bir eklentiye sahiptir.
Durma fonksiyonuna iliskin seçenekler sunulmakta, tekrar sayısı ile raporlama döngüsü
ayarlanabilmektedir. “Log” sekmesinden programın çalısması esnasında olusan islem
günlüklerine ulasılabilmektedir. “Graphic” sekmesinden eğitim ve test sonrası uyum grafiklerine ulasılabilmekte, “Fine Tuning” ve “Cascade Tuning” ile kademeli
özellestirmeler gerçeklestirilebilmektedir

...

Read More

Yapay Sinir Ağları Metodu İle Kalıp İşlerinde Bir Verimlilik ve Adam-Saat Tahmini Modeli

Yapay Sinir Ağları Metodu İle Kalıp İşlerinde Bir Verimlilik ve Adam-Saat Tahmini Modeli

İSTANBUL KÜLTÜR ÜNİVERSİTESİ
FEN BİLİMLERİ ENSTİTÜSÜ
YÜKSEK LİSANS TEZİ - Eylül 2009
Murat SÖNMEZ

Özet :

1980’li yılların başından itibaren mühendislikte artarak uygulama alanı bulan
yapay sinir ağları yöntemi, temelinde insan beyninin çalışma ilkelerini taklit
ederek çalışan bir problem çözümleme yöntemidir. Yöntemin en önemli özelliği
gerçek veriler ile kurulan modelin eğitilmesi ve eğitilmiş olan modelin yeni
veriler için sonuç üretebilmesidir. Bu bağlamda kurulan model sürekli olarak
yeni veriler ile sürekli kendini yenileyebilmesidir. Diğer bir deyişle model
sürekli öğrenerek kendini geliştirebilmektedir. Bu çalışmada, bina türü
projelerde kaba yapı maliyetleri içerisinde önemli yer tutan kalıp işlerine ait
adam-saat ve verimlilik değerlerinin sağlıklı tahmini amacıyla yapay sinir ağları
yöntemi ile bir karar destek sistemi oluşturulması hedeflenmiştir. Bu amaçla
çalışmanın ilk aşamasında bir yapay sinir ağı oluşturulmuştur. Bu aşamanın en
önemli kısmı girdi ve çıktı değişkenlerinin tespitidir. İkinci aşamada
oluşturulan bu ağ elde mevcut bulunan üstyapı projelerine ait kalıp puantajları
eğitilmiştir. Üçüncü ve son aşamada ise modelin sağlıklı çalışıp çalışmadığı
farklı projelerden elde edilen veriler ile test edilmiştir.

FannTool Hakkında ;

...
Çalışmada Fanntool 1.0 versiyonlu yazılım kullanılmıştır. Windows tabanlı
ve açık kodlu çalışan bir yazılımdır. Ayrıca diğer işletim sistemleri içinde
versiyonları (Linux, Dos v.s) mevcuttur.
...
Analizler ve testler sırasında program üzerinde oluşan değerler bir .log dosyası ile kayıt altına alınabiliyor. Programın çalışma parametreleri, komutları ve bunun gibi diğer özellikleri açık kod sayesinde kullanıcıların da görebileceği şekilde tasarlanmıştır. Ayrıca program üzerinde yukarıdaki bölümlerde anlatılan çeşitli fonksiyonlar ve bunların değişik versiyonları mevcuttur. Bu  fonksiyonlardan herhangi birisi seçilip kolayca analiz yapılabilir. Yine program dahilinde normalizasyon işlemi gerçekleştirilebilir. Fanntool kullanıcıların ihtiyacına göre en iyi öğrenme algoritmasını ve transfer fonksiyolarını kendi içersindeki bir analiz yöntemi ile belirleyebilmektedir.
...

Read More

Large Scale Reinforcement Learning using Q-SARSA(λ) and Cascading Neural Networks

M.Sc. Thesis
Steffen Nissen
October 8, 2007
Department of Computer Science
University of Copenhagen
Denmark

Abstract

This thesis explores how the novel model-free reinforcement learning algorithm Q-SARSA(λ) can be combined with the constructive neural network training algorithm Cascade 2, and how this combination can scale to the large problem of backgammon.
In order for reinforcement learning to scale to larger problem sizes, it needs to be combined with a function approximator such as an artificial neural network. Reinforcement learning has traditionally been combined with simple incremental neural network training algorithms, but more advanced training algorithms like Cascade 2 exists that have the potential of achieving much higher performance. All of these advanced training algorithms are, however, batch algorithms and since reinforcement learning is incremental this poses a  challenge. As of now the potential of the advanced algorithms have not been fully exploited and the few  combinational methods that have been tested have failed to produce a solution that can scale to larger  problems.
The standard reinforcement learning algorithms used in combination with neural networks are Q(λ) and SARSA(λ), which for this thesis have been combined to form the Q-SARSA(λ) algorithm. This algorithm has been combined with the Cascade 2 neural network training algorithm, which is especially interesting because it is a constructive algorithm that can grow a neural network by gradually adding neurons. For combining Cascade 2 and Q-SARSA(λ) two new methods have been developed: The NFQ-SARSA(λ) algorithm, which is an enhanced version of Neural Fitted Q Iteration and the novel sliding window cache.
The sliding window cache and Cascade 2 are tested on the medium sized mountain car and cart pole problems and the large backgammon problem. The results from the test show that Q-SARSA(λ) performs better than Q(λ) and SARSA(λ) and that the sliding window cache in combination with Cascade 2 and Q-SARSA(λ) performs significantly better than incrementally trained reinforcement learning. For the cart pole problem the algorithm performs especially well and learns a policy that can balance the pole for the complete 300 steps after only 300 episodes of learning, and its resulting neural network contains only one hidden neuron. This should be compared to 262 steps for the incremental algorithm after 10,000 episodes of
learning. The sliding window cache scales well to the large backgammon problem and wins 78% of the games against a heuristic player, while incremental training only wins 73% of the games. The  FQ-SARSA(λ) algorithm also outperforms the incremental algorithm for the medium sized problems, but it is not able to scale to backgammon.
The sliding window cache in combination with Cascade 2 and Q-SARSA(λ) performs better than incrementally trained reinforcement learning for both medium sized and large problems and it is the first combination of advanced neural network training algorithms and reinforcement learning that can scale to larger problems.

 Large Scale Reinforcement Learning using Q-SARSA(λ) and Cascading Neural Networks

Read More

Implementation of a Fast Artificial Neural Network Library

Steffen Nissen
October 31, 2003
Department of Computer Science
University of Copenhagen (DIKU)

Abstract

This report describes the implementation of a fast arti ficial neural network library in ANSI C called fann. The library implements multilayer feedforward networks with support for both fully connected and sparse connected networks. Fann off ers support for execution in fixed point arithmetic to allow for fast execution on systems with no  floating point processor. To overcome the problems of integer overflow, the library calculates a position of the decimal point after training and guarantees that integer overflow can not occur with this decimal point.
The library is designed to be fast, versatile and easy to use. Several benchmarks have been executed to test the performance of the library. The results show that the fann library is signi cantly faster than other libraries on systems without a  floating point processor, while the performance was comparable to other highly optimized libraries on systems with a floating point processor.

Implementation of a Fast Artificial Neural Network Library (fann)

Read More