SlideShare uma empresa Scribd logo
1 de 9
CprE 491 Project Plan:
 Recurrent Neural Networks
         Dec01-06




     Team Members:
      Andrew Brown
        David Doty
      Corey Folkerts
      Michael Kuehl


      Client / Advisor:
      Dr. Eric Bartlett
Table of Contents

List of Figures
Figure 1 – Gantt Chart of Timeline for Project Plan, pg. 8

List of Tables
Table 1 – Financial Budget, pg. 8
Table 2 – Personal Effort Budget, pg.8
Abstract
The general problem at hand is that of supervised learning. This is to say, can computers
learn by example? Specifically, this problem entails learning from data that may not
have a pre-specified form, rendering it difficult to solve through classical statistical
methods. In addition, many non-classical methods, such as feed-forward neural nets and
decision trees, encounter difficulty in learning from data that does not follow an
[obvious] pattern. In this project, this problem is attacked with a recurrent neural
network. This is a simplified model inspired by the human brain, consisting of models of
neurons connected to each other in an arbitrary fashion. While the flexibility of this
architecture makes general-purpose algorithms for analyzing it difficult to develop, it is
hoped that this flexible architecture will also allow recurrent neural net to solve more
complex learning problems than traditional approaches.

Acknowledgement
Iowa State’s Adaptive Computing Laboratory (ACL) will provide much of the
framework for the recurrent neural net. In particular, much of the knowledge of the
technical details of neural net operation is owed to Dr. Eric Bartlett.
Definition of Terms
ACL – Iowa State’s Adaptive Computing Laboratory

Artificial Neural Net – A simple model of learning consisting of two or more models of
neurons connected to each other, as a graph.

Feed-Forward Net – A neural net architecture in which the neurons of the network are
connected in a multi-partite structure, with the neurons divided into layers. The layers of
neurons, besides the last layer containing the output neuron, are known as hidden layers.

Network Model - The way the neurons of the network are connected (the architecture).

Neuron Model – A simple model of a biological neuron. It takes multiple inputs,
multiplies each of them by a separate weight, sums the products together, and outputs a
transfer function of this sum. The transfer function determines the neuron model.

Recurrent Net – A neural network architecture in which every neuron may be connected
to every other neuron. The output neuron is chosen arbitrarily.

Toolkit – The application developed by the ACL consisting of a collection of neural
network data modeling programs.
Introduction
  •General background - The machine learning problem is one of the oldest, and still
  one of the most daunting, in computer science. In general, can a computer, given
  examples of input data and output data, learn to predict future output given only the
  input? The applications of a solution to this general problem are nearly without limit.
  Given recorded weather patterns from previous days, can a computer predict the
  weather for the next day? Given pictures of a person from different angles, and with
  different facial expressions, can a computer pick that person out of a picture of a
  crowd? Given enough time to play Gary Kasparov in chess, could a computer learn
  to improve its performance sufficiently to beat him, without any previous knowledge
  of chess strategy? Traditionally, the learning problem has been analyzed through
  classical statistical methods. These methods, however, are limited in their usefulness.
  Consequently, researchers have turned to other approaches. One approach, neural
  networks, is inspired by the human brain’s ability to learn seemingly complex
  problems, such as facial recognition, with comparatively limited data. The Adaptive
  Computing Laboratory’s (ACL) Neural Toolkit currently contains a neural network
  architecture known as a feed-forward neural net. The goal of this project is to
  implement a recurrent neural network for use in the Toolkit. Simplistically, the main
  difference between a feed-forward net and a recurrent net is complexity. Feed-
  forward nets are simpler and easier to analyze, but the complexity of problems they
  can solve is limited. Recurrent nets mimic more closely the design of the brain and
  may allow computers to solve more challenging problems, yet are more difficult to
  analyze and have more inefficient learning algorithms.
  •Technical problem - The recurrent neural net will be implemented in software. An
  algorithm will need to be chosen to train the neural net. An algorithm will also need
  to be chosen to determine the output of the neural net. These algorithms must then be
  implemented in C++. Once these objectives have been accomplished, the network
  will be integrated into the existing Toolkit framework using a common user interface.
  •Operating environment - The operating environment is the Toolkit. The network
  will depend on the Toolkit to format input and output data and to display results. The
  Toolkit runs on the Microsoft Windows 95/98/2000/NT 4.0 operating system.
  •Intended user(s) and use(s) - The primary user of the Toolkit is the ACL. Other
  users might include clients of the ACL. Uses include learning problems in which
  other learning techniques have been shown to be ineffective. Experiments will need
  to be done to determine for what uses a recurrent neural net is more effective than
  other methods.
  •Assumptions and limitations
      •Assumptions:
          •Users of this module will already know how to use the Toolkit.
          •There is a plausible method to do this and we will figure out how to do it.
          •There is some practical use in doing this.
      •Limitations:
          •We have five months to work on the project.
Design Requirements
   •Design objectives
      • Develop a neuron model - This is the transfer function of the neurons.
      • Develop a network model - This dictates the interconnection of neurons.
      • Develop training algorithm - This algorithm will adjust the weights of the
          neurons to minimize the cost function.
   •Functional requirements
      • Be callable from a system call. This allows the Toolkit to run it.
      • Implement I/O interfaces - Formats of input and output files should match
          those of the Toolkit.
      • Should terminate on a definite stopping condition.
      • Should report progress to the user so that the user may decide intelligently
          whether to stop it before the stopping condition is reached.
   •Design constraints -
      • Uses only Windows system calls
      • Works only with normalized data
      • Runs with the minimum specifications of the Toolkit
   •Measurable milestones
      • Have a preliminary neuron model and network model
      • Complete testing on network model
      • Have a preliminary training algorithm
      • Complete testing of training algorithm on various data sets
      • Incorporate into the toolkit

End-Product Description
We have implemented a recurrent neural network module that integrates with the existing
ACL Neural Network Toolkit. It analyzes normalized training sets to infer dependencies
between the input and output of the training set. It then uses these dependencies to predict
output given arbitrary inputs.
Approach and Design
  •Technical approaches - One possible approach is the Rapid Prototype Model of
  software development. In this model, we design a prototype recurrent network and
  analyze its performance. The design goes through many iterations of design and
  analysis before moving to final implementation. Another possible approach is the
  Waterfall Model. In this model we break the design into five sections: requirements
  analysis, general design, detailed design, coding, and testing. The project proceeds
  linearly through the five phases. The following criteria will be used in deciding
  between these two: flexible design schedule, ability to provide measurable results in a
  timely manner, minimization of overhead.
  •Technical design - The following items must be defined:
       •Input format – This is the format of the data to input into the system.
       •Output format – This is the format of the data to output from the system.
       •Data structures – These will hold information about the neural net.
       •Evaluation algorithm – This will calculate the output from the neural net.
       •Cost functions – These will calculate the error of the model over the data.
       •Training algorithm – This will develop the model from the data.
  •Testing description - The effectiveness of the project results will be measured by the
  performance of the recurrent neural net. This performance will be measured using
  common cost functions, and these results will be compared to the performance of
  traditional methods. Computational efficiency will not be a significant factor in
  measuring performance, for this reason: The recurrent neural net is not expected to
  beat the performance of methods that are already in place. The purpose of the project
  is to lay a framework from which future efficiency- and performance-enhancing
  modifications to the recurrent neural net may easily be added.
  •Risks and risk management - Some of the risks we may potentially encounter are
  inability of our chosen development model to deliver a working recurrent neural
  network on time, loss of a team member, failure of our developed product to meet our
  expectations for performance, and an inability of our product to reach a solution to a
  test problem. Proper documentation will ensure that future ACL researchers will be
  able to continue the work easily and hopefully find a solution.
Financial Budget
Table 1 Financial budget
                                           Item                                         Original Estimated Cost
                                           Poster                                       $50.00
                                           Total estimated cost                         $50.00




Table 2 Personnel effort budget
                                           Personnel                                   Original Estimated Effort
                                           David Doty                                  159 hours
                                           Michael Kuehl                               154 hours
                                           Corey Folkerts                              157 hours
                                           Andrew Brown                                163 hours
                                           Total estimated effort                      615 hours

Project Schedule
                                         Feb 11, '01    Feb 18, '01 Feb 25, '01 Mar 4, '01 Mar 11, '01 Mar 18, '01 Mar 25, '01 Apr 1, '01 Apr 8, '01 Apr 15, '01
ID                Task Name                      TWT F SSMTWT F SSMTWT F S SMTWT F S SMTWT F S SMTWT F S SMTWT F SS MTWT F S S MTWT F S SMTW
 1 Project Poster



 2 Design Preliminary Network Structure



 3 Test Preliminary Network Structure



 4 Decide on Network Structure



 5 Design Preliminary Training Algorithm



 6 Test Preliminary Training Algorithm



 7 Decide on Training Algorithm



 8 Design Report



 9 Research Toolkit Implementation




Figure 1 – Gantt Chart of Timeline for Project Plan
Project Team Information:
   David Doty
   4541 Steinbeck, #2 Ames, IA, 50014
   515-292-8261
   ddoty@iastate.edu
   CprE

   Corey Folkerts
   Friley 3349 Noble, Ames, IA, 50012
   515-572-5519
   gumby145@iastate.edu
   CprE

   Michael Kuehl
   4541 Steinbeck, #2 Ames, IA, 50014
   515-292-8261
   mkuehl@iastate.edu
   CprE

   Andrew Brown
   610 7th St., Ames, IA, 50010
   515-233-3465
   ajbrown@iastate.edu
   CprE

Summary
Because recurrent neural networks are able to solve problems of such a type or
complexity that other methods have proved insufficient or infeasible this project will
offer a necessary addition to the diversity of the ACL's Neural Toolkit. By clearly
defining the goals of this project and scheduling such that we continuously make progress
towards those goals via the milestones defined it is the belief that this project will come
to fruition.


References
Eric B Bartlett, Associate Professor
Department of Electrical and Computer Engineering
Iowa State University

Machine Learning, Tom Mitchell, McGraw-Hill, 1997

Mais conteúdo relacionado

Mais procurados

NEURAL Network Design Training
NEURAL Network Design  TrainingNEURAL Network Design  Training
NEURAL Network Design Training
ESCOM
 
Neural network
Neural networkNeural network
Neural network
Saddam Hussain
 
Forecasting of Sales using Neural network techniques
Forecasting of Sales using Neural network techniquesForecasting of Sales using Neural network techniques
Forecasting of Sales using Neural network techniques
Hitesh Dua
 
Learning to learn unlearned feature for segmentation
Learning to learn unlearned feature for segmentationLearning to learn unlearned feature for segmentation
Learning to learn unlearned feature for segmentation
NAVER Engineering
 
Task Adaptive Neural Network Search with Meta-Contrastive Learning
Task Adaptive Neural Network Search with Meta-Contrastive LearningTask Adaptive Neural Network Search with Meta-Contrastive Learning
Task Adaptive Neural Network Search with Meta-Contrastive Learning
MLAI2
 

Mais procurados (20)

NEURAL Network Design Training
NEURAL Network Design  TrainingNEURAL Network Design  Training
NEURAL Network Design Training
 
Advance deep learning
Advance deep learningAdvance deep learning
Advance deep learning
 
Cost-effective Interactive Attention Learning with Neural Attention Process
Cost-effective Interactive Attention Learning with Neural Attention ProcessCost-effective Interactive Attention Learning with Neural Attention Process
Cost-effective Interactive Attention Learning with Neural Attention Process
 
Neural network
Neural networkNeural network
Neural network
 
Artificial Neural Network and its Applications
Artificial Neural Network and its ApplicationsArtificial Neural Network and its Applications
Artificial Neural Network and its Applications
 
Neural network image recognition
Neural network image recognitionNeural network image recognition
Neural network image recognition
 
Soft Computing
Soft ComputingSoft Computing
Soft Computing
 
Regression and Artificial Neural Network in R
Regression and Artificial Neural Network in RRegression and Artificial Neural Network in R
Regression and Artificial Neural Network in R
 
Forecasting of Sales using Neural network techniques
Forecasting of Sales using Neural network techniquesForecasting of Sales using Neural network techniques
Forecasting of Sales using Neural network techniques
 
Complex system
Complex systemComplex system
Complex system
 
A Defect Prediction Model for Software Product based on ANFIS
A Defect Prediction Model for Software Product based on ANFISA Defect Prediction Model for Software Product based on ANFIS
A Defect Prediction Model for Software Product based on ANFIS
 
neuralAC
neuralACneuralAC
neuralAC
 
Neural Networks in Data Mining - “An Overview”
Neural Networks  in Data Mining -   “An Overview”Neural Networks  in Data Mining -   “An Overview”
Neural Networks in Data Mining - “An Overview”
 
Face Recognition: From Scratch To Hatch / Эдуард Тянтов (Mail.ru Group)
Face Recognition: From Scratch To Hatch / Эдуард Тянтов (Mail.ru Group)Face Recognition: From Scratch To Hatch / Эдуард Тянтов (Mail.ru Group)
Face Recognition: From Scratch To Hatch / Эдуард Тянтов (Mail.ru Group)
 
Learning to learn unlearned feature for segmentation
Learning to learn unlearned feature for segmentationLearning to learn unlearned feature for segmentation
Learning to learn unlearned feature for segmentation
 
C3 w5
C3 w5C3 w5
C3 w5
 
Expert system neural fuzzy system
Expert system neural fuzzy systemExpert system neural fuzzy system
Expert system neural fuzzy system
 
Task Adaptive Neural Network Search with Meta-Contrastive Learning
Task Adaptive Neural Network Search with Meta-Contrastive LearningTask Adaptive Neural Network Search with Meta-Contrastive Learning
Task Adaptive Neural Network Search with Meta-Contrastive Learning
 
Ai and neural networks
Ai and neural networksAi and neural networks
Ai and neural networks
 
Introduction to artificial neural network
Introduction to artificial neural networkIntroduction to artificial neural network
Introduction to artificial neural network
 

Destaque

Angel and Bridge
Angel and BridgeAngel and Bridge
Angel and Bridge
butest
 
4/22/02
4/22/024/22/02
4/22/02
butest
 
Word version - Home - Lethbridge College
Word version - Home - Lethbridge CollegeWord version - Home - Lethbridge College
Word version - Home - Lethbridge College
butest
 
Report
ReportReport
Report
butest
 
The Machine That Could Learn Anything
The Machine That Could Learn AnythingThe Machine That Could Learn Anything
The Machine That Could Learn Anything
butest
 
LESSONS FROM THE MICHAEL JACKSON TRIAL
LESSONS FROM THE MICHAEL JACKSON TRIALLESSONS FROM THE MICHAEL JACKSON TRIAL
LESSONS FROM THE MICHAEL JACKSON TRIAL
butest
 
1. MPEG I.B.P frame之不同
1. MPEG I.B.P frame之不同1. MPEG I.B.P frame之不同
1. MPEG I.B.P frame之不同
butest
 

Destaque (9)

Angel and Bridge
Angel and BridgeAngel and Bridge
Angel and Bridge
 
ppt
pptppt
ppt
 
4/22/02
4/22/024/22/02
4/22/02
 
Word version - Home - Lethbridge College
Word version - Home - Lethbridge CollegeWord version - Home - Lethbridge College
Word version - Home - Lethbridge College
 
Report
ReportReport
Report
 
The Machine That Could Learn Anything
The Machine That Could Learn AnythingThe Machine That Could Learn Anything
The Machine That Could Learn Anything
 
BLPs
BLPsBLPs
BLPs
 
LESSONS FROM THE MICHAEL JACKSON TRIAL
LESSONS FROM THE MICHAEL JACKSON TRIALLESSONS FROM THE MICHAEL JACKSON TRIAL
LESSONS FROM THE MICHAEL JACKSON TRIAL
 
1. MPEG I.B.P frame之不同
1. MPEG I.B.P frame之不同1. MPEG I.B.P frame之不同
1. MPEG I.B.P frame之不同
 

Semelhante a Table of Contents

ExaLearn Overview - ECP Co-Design Center for Machine Learning
ExaLearn Overview - ECP Co-Design Center for Machine LearningExaLearn Overview - ECP Co-Design Center for Machine Learning
ExaLearn Overview - ECP Co-Design Center for Machine Learning
inside-BigData.com
 
A Survey of Machine Learning Methods Applied to Computer ...
A Survey of Machine Learning Methods Applied to Computer ...A Survey of Machine Learning Methods Applied to Computer ...
A Survey of Machine Learning Methods Applied to Computer ...
butest
 

Semelhante a Table of Contents (20)

Deep learning summary
Deep learning summaryDeep learning summary
Deep learning summary
 
IRJET- Deep Learning Model to Predict Hardware Performance
IRJET- Deep Learning Model to Predict Hardware PerformanceIRJET- Deep Learning Model to Predict Hardware Performance
IRJET- Deep Learning Model to Predict Hardware Performance
 
IRJET- Analysis of PV Fed Vector Controlled Induction Motor Drive
IRJET- Analysis of PV Fed Vector Controlled Induction Motor DriveIRJET- Analysis of PV Fed Vector Controlled Induction Motor Drive
IRJET- Analysis of PV Fed Vector Controlled Induction Motor Drive
 
IRJET-Performance Enhancement in Machine Learning System using Hybrid Bee Col...
IRJET-Performance Enhancement in Machine Learning System using Hybrid Bee Col...IRJET-Performance Enhancement in Machine Learning System using Hybrid Bee Col...
IRJET-Performance Enhancement in Machine Learning System using Hybrid Bee Col...
 
Blood Cell Image Classification for Detecting Malaria using CNN
Blood Cell Image Classification for Detecting Malaria using CNNBlood Cell Image Classification for Detecting Malaria using CNN
Blood Cell Image Classification for Detecting Malaria using CNN
 
Application of machine learning and cognitive computing in intrusion detectio...
Application of machine learning and cognitive computing in intrusion detectio...Application of machine learning and cognitive computing in intrusion detectio...
Application of machine learning and cognitive computing in intrusion detectio...
 
DATA SCIENCE
DATA SCIENCEDATA SCIENCE
DATA SCIENCE
 
Metaphorical Analysis of diseases in Tomato leaves using Deep Learning Algori...
Metaphorical Analysis of diseases in Tomato leaves using Deep Learning Algori...Metaphorical Analysis of diseases in Tomato leaves using Deep Learning Algori...
Metaphorical Analysis of diseases in Tomato leaves using Deep Learning Algori...
 
EMPIRICAL APPLICATION OF SIMULATED ANNEALING USING OBJECT-ORIENTED METRICS TO...
EMPIRICAL APPLICATION OF SIMULATED ANNEALING USING OBJECT-ORIENTED METRICS TO...EMPIRICAL APPLICATION OF SIMULATED ANNEALING USING OBJECT-ORIENTED METRICS TO...
EMPIRICAL APPLICATION OF SIMULATED ANNEALING USING OBJECT-ORIENTED METRICS TO...
 
IRJET-AI Neural Network Disaster Recovery Cloud Operations Systems
IRJET-AI Neural Network Disaster Recovery Cloud Operations SystemsIRJET-AI Neural Network Disaster Recovery Cloud Operations Systems
IRJET-AI Neural Network Disaster Recovery Cloud Operations Systems
 
Activity Monitoring Using Wearable Sensors and Smart Phone
Activity Monitoring Using Wearable Sensors and Smart PhoneActivity Monitoring Using Wearable Sensors and Smart Phone
Activity Monitoring Using Wearable Sensors and Smart Phone
 
ExaLearn Overview - ECP Co-Design Center for Machine Learning
ExaLearn Overview - ECP Co-Design Center for Machine LearningExaLearn Overview - ECP Co-Design Center for Machine Learning
ExaLearn Overview - ECP Co-Design Center for Machine Learning
 
Comparative Study of Pre-Trained Neural Network Models in Detection of Glaucoma
Comparative Study of Pre-Trained Neural Network Models in Detection of GlaucomaComparative Study of Pre-Trained Neural Network Models in Detection of Glaucoma
Comparative Study of Pre-Trained Neural Network Models in Detection of Glaucoma
 
Artificial Neural Network: A brief study
Artificial Neural Network: A brief studyArtificial Neural Network: A brief study
Artificial Neural Network: A brief study
 
A Survey of Machine Learning Methods Applied to Computer ...
A Survey of Machine Learning Methods Applied to Computer ...A Survey of Machine Learning Methods Applied to Computer ...
A Survey of Machine Learning Methods Applied to Computer ...
 
Using Deep Learning on Apache Spark to Diagnose Thoracic Pathology from Chest...
Using Deep Learning on Apache Spark to Diagnose Thoracic Pathology from Chest...Using Deep Learning on Apache Spark to Diagnose Thoracic Pathology from Chest...
Using Deep Learning on Apache Spark to Diagnose Thoracic Pathology from Chest...
 
11_Saloni Malhotra_SummerTraining_PPT.pptx
11_Saloni Malhotra_SummerTraining_PPT.pptx11_Saloni Malhotra_SummerTraining_PPT.pptx
11_Saloni Malhotra_SummerTraining_PPT.pptx
 
Artificial Neural Networks for data mining
Artificial Neural Networks for data miningArtificial Neural Networks for data mining
Artificial Neural Networks for data mining
 
Artificial Neural Networks for Data Mining
Artificial Neural Networks for Data MiningArtificial Neural Networks for Data Mining
Artificial Neural Networks for Data Mining
 
Presentation 7.pptx
Presentation 7.pptxPresentation 7.pptx
Presentation 7.pptx
 

Mais de butest

EL MODELO DE NEGOCIO DE YOUTUBE
EL MODELO DE NEGOCIO DE YOUTUBEEL MODELO DE NEGOCIO DE YOUTUBE
EL MODELO DE NEGOCIO DE YOUTUBE
butest
 
Timeline: The Life of Michael Jackson
Timeline: The Life of Michael JacksonTimeline: The Life of Michael Jackson
Timeline: The Life of Michael Jackson
butest
 
Popular Reading Last Updated April 1, 2010 Adams, Lorraine The ...
Popular Reading Last Updated April 1, 2010 Adams, Lorraine The ...Popular Reading Last Updated April 1, 2010 Adams, Lorraine The ...
Popular Reading Last Updated April 1, 2010 Adams, Lorraine The ...
butest
 
LESSONS FROM THE MICHAEL JACKSON TRIAL
LESSONS FROM THE MICHAEL JACKSON TRIALLESSONS FROM THE MICHAEL JACKSON TRIAL
LESSONS FROM THE MICHAEL JACKSON TRIAL
butest
 
Com 380, Summer II
Com 380, Summer IICom 380, Summer II
Com 380, Summer II
butest
 
The MYnstrel Free Press Volume 2: Economic Struggles, Meet Jazz
The MYnstrel Free Press Volume 2: Economic Struggles, Meet JazzThe MYnstrel Free Press Volume 2: Economic Struggles, Meet Jazz
The MYnstrel Free Press Volume 2: Economic Struggles, Meet Jazz
butest
 
MICHAEL JACKSON.doc
MICHAEL JACKSON.docMICHAEL JACKSON.doc
MICHAEL JACKSON.doc
butest
 
Social Networks: Twitter Facebook SL - Slide 1
Social Networks: Twitter Facebook SL - Slide 1Social Networks: Twitter Facebook SL - Slide 1
Social Networks: Twitter Facebook SL - Slide 1
butest
 
Facebook
Facebook Facebook
Facebook
butest
 
Executive Summary Hare Chevrolet is a General Motors dealership ...
Executive Summary Hare Chevrolet is a General Motors dealership ...Executive Summary Hare Chevrolet is a General Motors dealership ...
Executive Summary Hare Chevrolet is a General Motors dealership ...
butest
 
Welcome to the Dougherty County Public Library's Facebook and ...
Welcome to the Dougherty County Public Library's Facebook and ...Welcome to the Dougherty County Public Library's Facebook and ...
Welcome to the Dougherty County Public Library's Facebook and ...
butest
 
NEWS ANNOUNCEMENT
NEWS ANNOUNCEMENTNEWS ANNOUNCEMENT
NEWS ANNOUNCEMENT
butest
 
C-2100 Ultra Zoom.doc
C-2100 Ultra Zoom.docC-2100 Ultra Zoom.doc
C-2100 Ultra Zoom.doc
butest
 
MAC Printing on ITS Printers.doc.doc
MAC Printing on ITS Printers.doc.docMAC Printing on ITS Printers.doc.doc
MAC Printing on ITS Printers.doc.doc
butest
 
Mac OS X Guide.doc
Mac OS X Guide.docMac OS X Guide.doc
Mac OS X Guide.doc
butest
 
WEB DESIGN!
WEB DESIGN!WEB DESIGN!
WEB DESIGN!
butest
 
Download
DownloadDownload
Download
butest
 
resume.doc
resume.docresume.doc
resume.doc
butest
 

Mais de butest (20)

EL MODELO DE NEGOCIO DE YOUTUBE
EL MODELO DE NEGOCIO DE YOUTUBEEL MODELO DE NEGOCIO DE YOUTUBE
EL MODELO DE NEGOCIO DE YOUTUBE
 
Timeline: The Life of Michael Jackson
Timeline: The Life of Michael JacksonTimeline: The Life of Michael Jackson
Timeline: The Life of Michael Jackson
 
Popular Reading Last Updated April 1, 2010 Adams, Lorraine The ...
Popular Reading Last Updated April 1, 2010 Adams, Lorraine The ...Popular Reading Last Updated April 1, 2010 Adams, Lorraine The ...
Popular Reading Last Updated April 1, 2010 Adams, Lorraine The ...
 
LESSONS FROM THE MICHAEL JACKSON TRIAL
LESSONS FROM THE MICHAEL JACKSON TRIALLESSONS FROM THE MICHAEL JACKSON TRIAL
LESSONS FROM THE MICHAEL JACKSON TRIAL
 
Com 380, Summer II
Com 380, Summer IICom 380, Summer II
Com 380, Summer II
 
PPT
PPTPPT
PPT
 
The MYnstrel Free Press Volume 2: Economic Struggles, Meet Jazz
The MYnstrel Free Press Volume 2: Economic Struggles, Meet JazzThe MYnstrel Free Press Volume 2: Economic Struggles, Meet Jazz
The MYnstrel Free Press Volume 2: Economic Struggles, Meet Jazz
 
MICHAEL JACKSON.doc
MICHAEL JACKSON.docMICHAEL JACKSON.doc
MICHAEL JACKSON.doc
 
Social Networks: Twitter Facebook SL - Slide 1
Social Networks: Twitter Facebook SL - Slide 1Social Networks: Twitter Facebook SL - Slide 1
Social Networks: Twitter Facebook SL - Slide 1
 
Facebook
Facebook Facebook
Facebook
 
Executive Summary Hare Chevrolet is a General Motors dealership ...
Executive Summary Hare Chevrolet is a General Motors dealership ...Executive Summary Hare Chevrolet is a General Motors dealership ...
Executive Summary Hare Chevrolet is a General Motors dealership ...
 
Welcome to the Dougherty County Public Library's Facebook and ...
Welcome to the Dougherty County Public Library's Facebook and ...Welcome to the Dougherty County Public Library's Facebook and ...
Welcome to the Dougherty County Public Library's Facebook and ...
 
NEWS ANNOUNCEMENT
NEWS ANNOUNCEMENTNEWS ANNOUNCEMENT
NEWS ANNOUNCEMENT
 
C-2100 Ultra Zoom.doc
C-2100 Ultra Zoom.docC-2100 Ultra Zoom.doc
C-2100 Ultra Zoom.doc
 
MAC Printing on ITS Printers.doc.doc
MAC Printing on ITS Printers.doc.docMAC Printing on ITS Printers.doc.doc
MAC Printing on ITS Printers.doc.doc
 
Mac OS X Guide.doc
Mac OS X Guide.docMac OS X Guide.doc
Mac OS X Guide.doc
 
hier
hierhier
hier
 
WEB DESIGN!
WEB DESIGN!WEB DESIGN!
WEB DESIGN!
 
Download
DownloadDownload
Download
 
resume.doc
resume.docresume.doc
resume.doc
 

Table of Contents

  • 1. CprE 491 Project Plan: Recurrent Neural Networks Dec01-06 Team Members: Andrew Brown David Doty Corey Folkerts Michael Kuehl Client / Advisor: Dr. Eric Bartlett
  • 2. Table of Contents List of Figures Figure 1 – Gantt Chart of Timeline for Project Plan, pg. 8 List of Tables Table 1 – Financial Budget, pg. 8 Table 2 – Personal Effort Budget, pg.8
  • 3. Abstract The general problem at hand is that of supervised learning. This is to say, can computers learn by example? Specifically, this problem entails learning from data that may not have a pre-specified form, rendering it difficult to solve through classical statistical methods. In addition, many non-classical methods, such as feed-forward neural nets and decision trees, encounter difficulty in learning from data that does not follow an [obvious] pattern. In this project, this problem is attacked with a recurrent neural network. This is a simplified model inspired by the human brain, consisting of models of neurons connected to each other in an arbitrary fashion. While the flexibility of this architecture makes general-purpose algorithms for analyzing it difficult to develop, it is hoped that this flexible architecture will also allow recurrent neural net to solve more complex learning problems than traditional approaches. Acknowledgement Iowa State’s Adaptive Computing Laboratory (ACL) will provide much of the framework for the recurrent neural net. In particular, much of the knowledge of the technical details of neural net operation is owed to Dr. Eric Bartlett.
  • 4. Definition of Terms ACL – Iowa State’s Adaptive Computing Laboratory Artificial Neural Net – A simple model of learning consisting of two or more models of neurons connected to each other, as a graph. Feed-Forward Net – A neural net architecture in which the neurons of the network are connected in a multi-partite structure, with the neurons divided into layers. The layers of neurons, besides the last layer containing the output neuron, are known as hidden layers. Network Model - The way the neurons of the network are connected (the architecture). Neuron Model – A simple model of a biological neuron. It takes multiple inputs, multiplies each of them by a separate weight, sums the products together, and outputs a transfer function of this sum. The transfer function determines the neuron model. Recurrent Net – A neural network architecture in which every neuron may be connected to every other neuron. The output neuron is chosen arbitrarily. Toolkit – The application developed by the ACL consisting of a collection of neural network data modeling programs.
  • 5. Introduction •General background - The machine learning problem is one of the oldest, and still one of the most daunting, in computer science. In general, can a computer, given examples of input data and output data, learn to predict future output given only the input? The applications of a solution to this general problem are nearly without limit. Given recorded weather patterns from previous days, can a computer predict the weather for the next day? Given pictures of a person from different angles, and with different facial expressions, can a computer pick that person out of a picture of a crowd? Given enough time to play Gary Kasparov in chess, could a computer learn to improve its performance sufficiently to beat him, without any previous knowledge of chess strategy? Traditionally, the learning problem has been analyzed through classical statistical methods. These methods, however, are limited in their usefulness. Consequently, researchers have turned to other approaches. One approach, neural networks, is inspired by the human brain’s ability to learn seemingly complex problems, such as facial recognition, with comparatively limited data. The Adaptive Computing Laboratory’s (ACL) Neural Toolkit currently contains a neural network architecture known as a feed-forward neural net. The goal of this project is to implement a recurrent neural network for use in the Toolkit. Simplistically, the main difference between a feed-forward net and a recurrent net is complexity. Feed- forward nets are simpler and easier to analyze, but the complexity of problems they can solve is limited. Recurrent nets mimic more closely the design of the brain and may allow computers to solve more challenging problems, yet are more difficult to analyze and have more inefficient learning algorithms. •Technical problem - The recurrent neural net will be implemented in software. An algorithm will need to be chosen to train the neural net. An algorithm will also need to be chosen to determine the output of the neural net. These algorithms must then be implemented in C++. Once these objectives have been accomplished, the network will be integrated into the existing Toolkit framework using a common user interface. •Operating environment - The operating environment is the Toolkit. The network will depend on the Toolkit to format input and output data and to display results. The Toolkit runs on the Microsoft Windows 95/98/2000/NT 4.0 operating system. •Intended user(s) and use(s) - The primary user of the Toolkit is the ACL. Other users might include clients of the ACL. Uses include learning problems in which other learning techniques have been shown to be ineffective. Experiments will need to be done to determine for what uses a recurrent neural net is more effective than other methods. •Assumptions and limitations •Assumptions: •Users of this module will already know how to use the Toolkit. •There is a plausible method to do this and we will figure out how to do it. •There is some practical use in doing this. •Limitations: •We have five months to work on the project.
  • 6. Design Requirements •Design objectives • Develop a neuron model - This is the transfer function of the neurons. • Develop a network model - This dictates the interconnection of neurons. • Develop training algorithm - This algorithm will adjust the weights of the neurons to minimize the cost function. •Functional requirements • Be callable from a system call. This allows the Toolkit to run it. • Implement I/O interfaces - Formats of input and output files should match those of the Toolkit. • Should terminate on a definite stopping condition. • Should report progress to the user so that the user may decide intelligently whether to stop it before the stopping condition is reached. •Design constraints - • Uses only Windows system calls • Works only with normalized data • Runs with the minimum specifications of the Toolkit •Measurable milestones • Have a preliminary neuron model and network model • Complete testing on network model • Have a preliminary training algorithm • Complete testing of training algorithm on various data sets • Incorporate into the toolkit End-Product Description We have implemented a recurrent neural network module that integrates with the existing ACL Neural Network Toolkit. It analyzes normalized training sets to infer dependencies between the input and output of the training set. It then uses these dependencies to predict output given arbitrary inputs.
  • 7. Approach and Design •Technical approaches - One possible approach is the Rapid Prototype Model of software development. In this model, we design a prototype recurrent network and analyze its performance. The design goes through many iterations of design and analysis before moving to final implementation. Another possible approach is the Waterfall Model. In this model we break the design into five sections: requirements analysis, general design, detailed design, coding, and testing. The project proceeds linearly through the five phases. The following criteria will be used in deciding between these two: flexible design schedule, ability to provide measurable results in a timely manner, minimization of overhead. •Technical design - The following items must be defined: •Input format – This is the format of the data to input into the system. •Output format – This is the format of the data to output from the system. •Data structures – These will hold information about the neural net. •Evaluation algorithm – This will calculate the output from the neural net. •Cost functions – These will calculate the error of the model over the data. •Training algorithm – This will develop the model from the data. •Testing description - The effectiveness of the project results will be measured by the performance of the recurrent neural net. This performance will be measured using common cost functions, and these results will be compared to the performance of traditional methods. Computational efficiency will not be a significant factor in measuring performance, for this reason: The recurrent neural net is not expected to beat the performance of methods that are already in place. The purpose of the project is to lay a framework from which future efficiency- and performance-enhancing modifications to the recurrent neural net may easily be added. •Risks and risk management - Some of the risks we may potentially encounter are inability of our chosen development model to deliver a working recurrent neural network on time, loss of a team member, failure of our developed product to meet our expectations for performance, and an inability of our product to reach a solution to a test problem. Proper documentation will ensure that future ACL researchers will be able to continue the work easily and hopefully find a solution.
  • 8. Financial Budget Table 1 Financial budget Item Original Estimated Cost Poster $50.00 Total estimated cost $50.00 Table 2 Personnel effort budget Personnel Original Estimated Effort David Doty 159 hours Michael Kuehl 154 hours Corey Folkerts 157 hours Andrew Brown 163 hours Total estimated effort 615 hours Project Schedule Feb 11, '01 Feb 18, '01 Feb 25, '01 Mar 4, '01 Mar 11, '01 Mar 18, '01 Mar 25, '01 Apr 1, '01 Apr 8, '01 Apr 15, '01 ID Task Name TWT F SSMTWT F SSMTWT F S SMTWT F S SMTWT F S SMTWT F S SMTWT F SS MTWT F S S MTWT F S SMTW 1 Project Poster 2 Design Preliminary Network Structure 3 Test Preliminary Network Structure 4 Decide on Network Structure 5 Design Preliminary Training Algorithm 6 Test Preliminary Training Algorithm 7 Decide on Training Algorithm 8 Design Report 9 Research Toolkit Implementation Figure 1 – Gantt Chart of Timeline for Project Plan
  • 9. Project Team Information: David Doty 4541 Steinbeck, #2 Ames, IA, 50014 515-292-8261 ddoty@iastate.edu CprE Corey Folkerts Friley 3349 Noble, Ames, IA, 50012 515-572-5519 gumby145@iastate.edu CprE Michael Kuehl 4541 Steinbeck, #2 Ames, IA, 50014 515-292-8261 mkuehl@iastate.edu CprE Andrew Brown 610 7th St., Ames, IA, 50010 515-233-3465 ajbrown@iastate.edu CprE Summary Because recurrent neural networks are able to solve problems of such a type or complexity that other methods have proved insufficient or infeasible this project will offer a necessary addition to the diversity of the ACL's Neural Toolkit. By clearly defining the goals of this project and scheduling such that we continuously make progress towards those goals via the milestones defined it is the belief that this project will come to fruition. References Eric B Bartlett, Associate Professor Department of Electrical and Computer Engineering Iowa State University Machine Learning, Tom Mitchell, McGraw-Hill, 1997