We will create python/svm_analysis.py, which will parse the POST data and request the respective SVM model from a mySQL database query. This query will be defined within a method from python/svm_model.py.

Note: The queried Support Vector Machine model, is previously defined by the same script python/svm_model.py, when imported and called from python/svm_training.py.

Note: It is important that we escape any parameters received from php/logic_loader.php.

For the remaining week, we have a temporary Windows 7 machine:

Windows edition
Windows 7 Professional
Copyright 2009 Microsoft Corporation. All rights reserved
Service Pack1

System
Manufacturer: Dell
Model: Precision M4600
Rating: 6.9 Windows Experience Index
Processor: Intel(R) Core(TM) i7-2760QM CPU @ 2.40GHz 2.40 GHz
Installed memory (RAM): 16.0 GB
System type: 64-bit Operating System
Pen and Touch: No Pen or Touch Input is available for the Display

Hard drive
File system: NTFS
Capacity: 237 GB

Unfortunately, we will not be setting up a dual-boot. Also, this machine already implements VMware Player 5.0.2 build-1031769. So, we will find the corresponding version of Ubuntu Server to work within the VMware Player.

Since the issue of setting up working environments has come up more than several times across other projects, it serves precursor for something fantastic. More specifically, we will research how to create a bootable USB drive.

We will research the practicality of the jquery autocomplete, by following a corresponding tutorial. This will be used to auto-suggest values for various input fields within the html form. These suggestions will be dynamically created based on the dataset chosen during training.

We will determine if we need a database scheme to store the data sets.

We will initially have html_form.js be responsible for adding additional form fields.

We want to ensure that we append additional form fields after the Remove buttons, not after the Add more buttons.

We no longer require the use of Flask. Therefore, we will remove the submodule, and any dependencies needed for Flask:

$ sudo apt-get remove libapache2-mod-wsgi

We will only display the form submit button for the following two cases:

• analysis session: known factors, independent variables have been provided
• training session: classification, dependent / independent variables are been provided

We will override the submit event using AJAX, and prevent the default action.

We will implement fieldsets to group similar form elements together in a <fieldset>.

Form elements added is always the second element in the field element array. We prefer elements be added to end of the field array. This will require adjusting html_form_delegator.js.

We need to install scikit-learn:

$ cd /var/www/html/machine-learning/scikit-learn/
$ python setup.py build
$ sudo python setup.py install

html_form.js will be responsible for creating additional form fieldsets based on various datalist choices. For example, the choice of Session Type appends the corresponding fieldset, that is either Training Session, or Analysis Session to the form DOM structure. Therefore, tests/php/index.php will be responsible for creating an initial form with just one fieldset:

• Session Type

Next, we will need to remove all form fieldsets within tests/php/index.php except the one listed above. Also, we need to add a reference to html_form.js, since it was earlier renamed to html_form_delegator.js.

We will create tests/index.php to be our test page for the Support Vector Machines (SVM). This page will pass a few parameters to the imported python/logic_loader.py. For example, we will pass the required parameter whether this is a training, or analysis instance (string), an optional object containing XML feed urls, an optional object containing attributes of XML elements we would like to parse for.

logic_loader.py will delegate the parameters to other python scripts in order to return either a message indicating a successful training session, or a prediction.

We will have the Known Factors fieldset determine the DOM creation of the Estimated Analysis fieldset. The changes required for this issue will be very similar to the issue Form: Display 'Training Type' Fieldset.

We will add the Fask submodule.

We will move the current source code for html_form.js into html_form_delegator.js. Then, we will need to update the reference of the rename javascript in tests/php/index.php.

Note: Once this is complete, html_form.js will create a single fieldset.

We will create python/logic_loader.php which will be the primary script that will import and instantiate other required python scripts. logic_loader.php may call upon the following scripts:

• xml_retriever.py
• data_creator.py
• svm_initializer.py

Note: this script will be passed three parameters:

• (required) type string, indicating whether this is a training, or analysis instance
• (optional) type object, containing XML feed urls
• (optional) type object, containing attributes of XML elements to be parsed

If both, optional parameters have not been passed in, and the required parameter indicates training this script will prompt users to manually input data (i.e. upload file). The latter option will be defined in this script. But, data_creator.py will not have the ability (yet) to parse the uploaded file.

We will add jquery-1.8.3.js from the whisper repository.

We will create python/xml_retriever.py as a generic class that will take two parameters:

• the URL of the xml document to parse
• an object containing the xml attributes to parse from the xml document

This class, after being instantiated will return an object. This return object will be the important xml attributes we wanted to parse for (as defined by the second parameter of this class).

We will create python/svm_training.py which will send data to python/data_creator.py, then store the SVM dataset into a mySQL database. Then, python/svm_training.py will call upon python/svm_model.py which will be responsible for creating the Support Vector Machine model, respectively.

Note: It is important that we escape any parameters received from php/logic_loader.php.

We will remove the multiple <br> tags generated by html_form.js.

Just as we have the <datalist> for the Session type, we need to create another <datalist> for Training Session Type. This field will need to be placed above the fieldset Supply Dataset within tests/php/index.php.

We will research which python web framework to implement for receiving POST data from an HTML5 <form> element, which we hope to send to a python script.

We need to document the steps required to install scikit-learn.

Note: We will install flask which will allow use to send POST data to the SVM component to the scikit-learn submodule.

To gain a better understanding of Support Vector Machines, we will research the following:

SVM Classification

• General Definition
• Requirement
• Other Modeling Techniques (SVM in general)
• Limitations (curse of dimensionality, applies generally to multi-dimensional models)

SVM Regression

• Background: Regression Analysis
• SVM: Regression Analysis

Loss Functions

• Loss Functions: Quick Summary
• Loss Functions: SVM Regression
• Loss Functions: SVM Classification
• Types of Regression

Now, that we have flask in this repository, we will proceed with the installation:

sudo apt-get install python-pip
sudo python setup.py develop

We need to create an initial outline for the README.md.

We will define the created python/data_creator.py, which will be responsible for storing our merged SVM dataset into a database.

Just as we have the <datalist> for the Session type, we need to create another <datalist> for Dataset Type. This field will need to be placed above the datalist Training Type, and Supply Dataset within tests/php/index.php.

This issue assists with the overall understanding of how Support Vector Machines are used. We will begin by choosing an acceptable XML feed to generate a data set.

We previously decided to have php as the language that distributes necessary parameters to our python scripts. Therefore, tests/php/index.php will need to be adjusted so that POST data is sent to python/logic_loader.php, not python/logic_loader.py.

Just as we have the <datalist> for the Session type, we need to create another <datalist> for Select Model. This field will need to be placed above the fieldset Known Factors within tests/php/index.php. The element will list all the trained models available for analysis.

We will add the following within <head> of tests/php/index.php:

<script src='src/js/jquery-1.8.3.js'></script>

We found the IRC #scikit-learn very useful. As such, we will use this implementation instead of the pyML we had previously considered. We may at a later time decide to use, or mirror the latter project to GitHub.

Currently, a choice on the Dataset Type fieldset creates both Supply Dataset, and Training Type fieldsets. Instead, we will add conditional javascript, when a user provides context for Supply Dataset, then the Training Type fieldset will be created in the DOM.

We will attempt to add pyML from SourceForge as a submodule in this GitHub repository. This will allow us to use the Support Vector Machines. However, this project doesn't exist in GitHub. We will contact the maintainer from SourceForge, Professor Asa Ben-Hur, and offer to mirror the code into GitHub.

The steps to mirror svn to GitHub is a fairly simple clone command.

IRC #cmusphinx (08/13/14 ~ 6:40pm EST):

jeffreylevesque: nshm, was it difficult to create a mirror from svn to github, like you did with sphinx libraries?

nshm: no, thats just few commands
nshm: git svn clone
nshm: git push
nshm: there are tutorials on the web
nshm: https://www.atlassian.com/git/migration#!migration-convert

We will create python/svm_initializer.py. This script will determine if we need to undergo training, or perform analysis. Therefore, depending on what is needed, the following may be called:

• svm_training.py
• svm_analysis.py

We will add a unique classname for each fieldset within the <form> element in tests/php/index.php.

We will need a function that counts the number of input fields created by the Add more button, and check whether all created field elements are defined. This function will be implemented for any fieldsets being created that depends on an input field that can be appended with additional elements. This will ensure that users provide data for all form fields, before succeeding fieldsets are available.

We will determine the syntax requirements for both Support Vector Machine Classification, and Regression within our implementation of scikit-learn.

SVM Classification

• Preprocessing Data Set
• Training On Data Set
• Implement new Classifier

SVM Regression

• Preprocessing Data Set
• Training On Data Set
• Implement new Regression Model

Since, we've added the Remove buttons, we now need to define its behavior. Upon being clicked, it must remove the last corresponding input element within the immediate containing fieldset.

We will add Remove buttons to the immediate right of the Add more buttons. This will be responsible for removing the last respective field element within the corresponding fieldset in the form.

We need to add the following scikit-learn dependency to the Linux Packages subsection:

$ sudo apt-get install python-numpy

We need to include instructions on how to use the Web Interface, and how to interpret prediction results.

The following needs to be marked up as code:

# General Packages:

# Scikit Package(s):
sudo apt-get install python-numpy

We need a more dynamic implementation of the SVM form tests/php/index.php.

We need to document the relevance of the python Flask submodule, and the steps required to install, and configure it within this project.

We need to describe our unit tests located in the /test/ directory. Then, we need to discuss the various logs this repository produces, and how it may be useful.

Users need to be able to add more than one uploaded file, more than one URL to an xml file for the SVM data set, and more than one dependent or independent variables.