neural network - What is the appropriate Machine Learning Algorithm for this scenario? -


i working on machine learning problem looks this:

input variables

 categorical         b     c     d   continuous     e

output variables

discrete(integers)     v     x     y  continuous     z

the major issue facing output variables not totally independent of each other , there no relation can established between them. is, there dependence not due causality (one value being high doesn't imply other high chances of other being higher improve)

an example be:

v - number of ad impressions

x - number of ad clicks

y - number of conversions

z - revenue

now, ad clicked, has first appear on search, click dependent on impression.

again, ad converted, has first clicked, again conversion dependent on click.

so running 4 instances of problem predicting each of output variables doesn't make sense me. infact there should way predict 4 taking care of implicit dependencies.

but can see, there won't direct relation, infact there probability involved can't worked out manually.

plus output variables not categorical in fact discrete , continuous.

any inputs on how go solving problem. guide me existing implementations same , toolkit use implement solution.

just random guess - think problem can targeted bayesian networks. think ?

bayesian networks fine in case. network won't huge either can live exact inference algorithms graph elimination or junction tree. if decide use bns, can use kevin murphy's bn toolbox. here link that. more general toolbox uses gibbs sampling approximate monte carlo inference, can use bugs.

edit:

as example @ famous sprinkler example here. totally discrete variables, define conditional probability tables in link. instance given today cloudy, there 0.8 probability of rain. define probability distributions, graph shows causality relations (i.e. if cloud rain etc.) query ask inference algorithm questions like, given grass wet; cloudy, raining, sprinkler on , on.

to use bns 1 needs system model described in terms of causality relations (directed acyclic graph) , probability transitions. if wanna learn system parameters there techniques em algorithm. however, learning graph structure hard task , supervised machine learning approaches better in case.


Comments

Post a Comment

Popular posts from this blog

java - Play! framework 2.0: How to display multiple image? -

i need display gallery of photos. here template: @(photos: list[photo]) @title = { <bold>gallery</bold> } @main(title,"photo"){ <ul class="thumbnails"> @for(photo <- photos) { <li class="span3"> <a href="#" class="thumbnail"> <img src="@photo.path" alt=""> </a> </li> } </ul> } and here controller method: public static result getphotos() { return ok(views.html.photo.gallery.render(photo.get())); } and here photo bean : @entity public class photo extends model { @id public long id; @required public string label; public string path; public photo(string path, string label) { this.path = path; this.label = label; } private static finder<long, photo> find = new finder<long, photo>( long.class, photo.class); public static list<phot...
Read more

gmail - Is there any documentation for read-only access to the Google Contacts API? -

i want read-only access user's google contacts. there stackoverflow member wanted this, , received following answer : yes, there is, use https://www.googleapis.com/auth/contacts.readonly scope , "view contacts". this exchange reference read-only access google contacts can find using google. simply using access token obtained attempt read usual google contacts api (i.e. https://www.google.com/m8/feeds ) fails insufficient permissions. similarly, using google contacts api on new target fails 404 errors (for example https://www.googleapis.com/auth/contacts.readonly/default/full ). is there documentation anywhere on how use api, or permissible way access google contacts api full read-write access?
Read more

php - Controller/JToolBar not working in Joomla 2.5 -

relevant code: jtoolbarhelper::custom('savecategories', 'save', '', 'save', false, false); ... <input type="hidden" name="controller" value="easyblogcontroller"> according can dig docs , own previous questions, should call savecategories() function in easyblogcontroller . i've tried setting value easyblog , easyblog.php (file name) current easyblogcontroller (class name). clicking on save button refreshes page. not redirect, echo or var_dump test code put in savecategories() function. var_dump ing jrequest::getvar('controller') , 'task' return correct values. creating controller object , using $controller->execute('task'); works. the values going depend on location of controller trying call, , have couple options here. typically there controller in base folder of component (likely components/com_easyblog) in file called controller.php , class name inside ...
Read more