Wed Aug 15, 2018 2:58 pm
Login Register Lost Password? Contact Us


Multilayer perceptron

Topics related to the set of Machine Learning libraries and Matrix processing algorithms

Thu Mar 19, 2015 12:11 am Change Time Zone

Hello,

I have this idea of implementing multi layer perceptron, I have just checked how perceptron is working. I need some guidance and inputs for starting with it. Please provide if pre requisites any are required, like if I need to know Roxie for it or can I start with only ecl.
Also please suggest if it can be done for Google summer competition, or proposals have to be submitted only for the listed topics.

Thanks,
Pooja.
chennapooja
 
Posts: 61
Joined: Wed Oct 08, 2014 11:49 pm

Thu Mar 19, 2015 1:58 pm Change Time Zone

We have a multilayer neural network. See ML.NeuralNetworks.ecl.

It is fairly new. So, you may not have it in your local copy of the ecl-ml repository. But, it is in the github repository.
tlhumphrey2
 
Posts: 250
Joined: Mon May 07, 2012 6:23 pm

Fri Mar 20, 2015 9:57 pm Change Time Zone

Dear Tlhumphrey,

Thanks for the information. I did not have that in my local copy. I will go through it and post if I face any confusions.

It will be grateful if any documents related to the algorithm development are shared.

Thanks and Regards,
Pooja.
chennapooja
 
Posts: 61
Joined: Wed Oct 08, 2014 11:49 pm

Mon Mar 23, 2015 3:16 pm Change Time Zone

Dear Team,

Please check if there is something wrong in below code as I am getting error "Error: Compile/Link failed for W20150322-004815 (see '//192.168.56.129/var/lib/HPCCSystems/myeclccserver/eclcc.log' for details) (0, 0), 3000, W20150322-004815"

I am trying to test neural networks - multi layer perceptron.
<code>IMPORT * FROM ML;
IMPORT ML.Mat;
//IMPORT ML.Tests.Explanatory as TE;

//This is the tennis-weather dataset transformed to discrete number values.
mlpRecord := RECORD
Types.t_RecordID id;
Types.t_FieldNumber layer_num;
Types.t_FieldNumber default;
Types.t_FieldNumber nodes;
END;

mlp_Data := DATASET([
{1,1,1,4},
{2,2,1,2},
{3,3,1,5}],
mlpRecord);

weatherRecord := RECORD
Types.t_RecordID id;
Types.t_FieldNumber outlook;
Types.t_FieldNumber temperature;
Types.t_FieldNumber humidity;
Types.t_FieldNumber windy;
Types.t_FieldNumber play;
END;

weather_Data := DATASET([
{1,0,0,1,0,0},
{2,0,0,1,1,0},
{3,1,0,1,0,1},
{4,2,1,1,0,1},
{5,2,2,0,0,1},
{6,2,2,0,1,0},
{7,1,2,0,1,1},
{8,0,1,1,0,0},
{9,0,2,0,0,1},
{10,2,1,0,0,1},
{11,0,1,0,1,1},
{12,1,1,1,1,1},
{13,1,0,0,0,1},
{14,2,1,1,1,0}],
weatherRecord);

indep_data1:= TABLE(weather_Data,{id, outlook, temperature, humidity, windy});
dep_data:= TABLE(weather_Data,{id, play});

ToField(indep_data1, pr_indep1);
indep1 := ML.Discretize.ByRounding(pr_indep1);
ToField(dep_data, pr_dep);
dep := ML.Discretize.ByRounding(pr_dep);

indep_data:= TABLE(mlp_Data,{id, layer_num, default, nodes});

ToField(indep_data, pr_indep);
indep := ML.Discretize.ByRounding(pr_indep);


trainer2:= ML.NeuralNetworks(indep); // Unpruned
model2:= trainer2.NNLearn(pr_indep1, pr_dep, trainer2.IntWeights,trainer2.IntBias);
results2:= trainer2.NNClassify(pr_indep1, model2);

//trainer2;
OUTPUT(model2, NAMED('Model2'));
//OUTPUT(SORT(trainer2.Model(model2), level, node_id), NAMED('DecTree_2'), ALL);
results2;</code>

Thanks and Regards,
Pooja.
chennapooja
 
Posts: 61
Joined: Wed Oct 08, 2014 11:49 pm

Fri Mar 27, 2015 10:40 pm Change Time Zone

Dear team,

Please throw me some light on above questions.

Thanks and Regards,
Pooja.
chennapooja
 
Posts: 61
Joined: Wed Oct 08, 2014 11:49 pm

Sun Mar 29, 2015 7:20 pm Change Time Zone

chennapooja,

Sorry for the late response. I haven't been around for a week.

You need to attach the eclcc.log if you want help with this specific problem. If you decide to go this route, also let me know if you can ssh into the THOR node where the eclcc.log file exists, i.e. 192.168.56.129.

But if you are just interested in testing ML.NeuralNetworks, look at the test case ML.Tests.Explanatory.NeuralNetworks_test.ecl.

If you are using ECL IDE, open the folder, ML/Tests/Explanatory, and right click on NeuralNetworks_test and select "Open in Builder Window". This puts NeuralNetworks_test in a builder window so you can execute it.

Then, you only have to "Submit" it to have it executed.

Also, you might want to compare NeuralNetworks_test to your code. You probably will learn something that will help you solve your problem.

Tim
tlhumphrey2
 
Posts: 250
Joined: Mon May 07, 2012 6:23 pm

Sun Mar 29, 2015 7:54 pm Change Time Zone

Thanks Tim, will check it out.
chennapooja
 
Posts: 61
Joined: Wed Oct 08, 2014 11:49 pm

Tue Mar 31, 2015 3:57 pm Change Time Zone

Dear Tim,

Even for the example NeuralNetworks_test.ecl, I am getting same error. To ssh the log, should I have some secure key or user id, password? How can I check the log, please update. Thanks in advance.

Regards,
Pooja.
chennapooja
 
Posts: 61
Joined: Wed Oct 08, 2014 11:49 pm

Mon Apr 06, 2015 6:11 pm Change Time Zone

Dear Team,

Any update here please. Thanks in advance.
Also I would like to know which kind of deep learning is implemented, Is it like Deep Belief Network or Boltzmann machine or convolutional networks etc...Are there any further proposals in the deep learning.

Thanks and Regards,
Pooja.
chennapooja
 
Posts: 61
Joined: Wed Oct 08, 2014 11:49 pm


Return to Machine Learning

Who is online

Users browsing this forum: No registered users and 1 guest

cron