Hi, I have taken only the Accuracy,Precision,F1 Score and Recall of neural networks. The values are listed below.
*Accuracy: 0.9712* *Precision: 0.9711317415957886* *Recall: 0.9708490075847962* *F1 Score: 0.9709903540085949* These are the hyper parameters user has set for the MNIST dataset for the UI. *Neural Network Type = Feed Forward Network* *Seed = 123* *Learning Rate = 0.006* *Batch Size = 128* *nEpoches = 15* *Iteration = 1* *Optimization Algorithm = Storchastic Gradient Descent* *Updater = Nesterovs* *Momentum = 0.9* *Pretrain = False* *Back Propargation = True* *Input Layer:* * Input Nodes = 784* *1st Hidden Layer:* * Number of Hidden Nodes = 1000* * WeightInit = Xavier* * Activation = RELU* *Output Layer:* * Output Nodes = 10* * WeightInit = Xavier* * Activation = Softmax* * Loss Function = Negetive Log Likelihood* For the training purposes and for the time being I have used the Dataset of MNIST which is available in DL4J library. *DataSetIterator trainIter = new MnistDataSetIterator(bachSize, true, (int)seed);* *DataSetIterator testIter = new MnistDataSetIterator(bachSize, false, (int) seed);* I will try to use the dataset from the WSO2 ML Server.I couldn't try it yesterday. The training time for the above scenario is 08 min 45s . I will share the code once I complete reading data set from the WSO2 ML Sever. Thank you. On Thu, Aug 11, 2016 at 8:43 AM, Upul Bandara <u...@wso2.com> wrote: > Sounds good > > Can you please share us the performance numbers (and confusion matrix) you > got with the MINIST? > Also, how did you do: > load MNIST data into WSO2 ML Server? > split training/testing subsets? > Tune hyper-parameters? > > What was the training time? > > Thanks, > Upul > > > On Thu, Aug 11, 2016 at 6:59 AM, Lakini Senanayaka < > lakinisenanayak...@gmail.com> wrote: > >> Hi. >> >> Thank you very much for sharing the code. >> I have solved many of the issues I had with the UI. >> >> I have tested the UI with MNIST dataset.It gave the correct results and >> the UI is working well with large datasets. >> >> The UI gets unresponsive for sometimes as the AJAX call takes nearly 5 >> minutes to give the response.Currently, I am handling that issue and I have >> to call the dataset from the ML server. >> >> I will share my documentation as soon as possible. >> >> Thank you. >> >> On Tue, Aug 9, 2016 at 9:52 AM, Upul Bandara <u...@wso2.com> wrote: >> >>> Following jag file ( please look at function loadDatasets(), function >>> updateDatasets(), function deleteDataset(datasetId) and etc) has everything >>> you need to know to get dataset from ML server to your UI. Please kindly >>> note that it is very difficult for us to give exact code snippet to meet >>> your requirement. As a student, you should be able to go through available >>> code which shows how to use APIs and uses those APIs in your applications. >>> >>> I think the API doc Supun shared with you and the following jag file >>> will help you to understand how to extract dataset and versions from the ML >>> Server and displaying those in your UI. >>> >>> BTW, did you manage to test your UI with large datasets such as MNIST. >>> MNIST is an image dataset but you can convert it to standard CSV format and >>> can be used to train NN using your UI. >>> >>> https://github.com/wso2/carbon-ml/blob/master/apps/ml/site/d >>> ata/datasets.jag >>> >>> On Mon, Aug 8, 2016 at 6:31 AM, Lakini Senanayaka < >>> lakinisenanayak...@gmail.com> wrote: >>> >>>> Hi, >>>> >>>> Thank you very much. >>>> >>>> But Upul told me about a code snippet which I can use inside the API >>>> implementation in the last meeting. >>>> >>>> However can I get the dataset file from calling *GET >>>> https://localhost:9443/api/datasets/{datasetId} >>>> <https://localhost:9443/api/datasets/%7BdatasetId%7D> * inside the API >>>> implementation in the backend (inside carbon-ml\components\ml\org.ws >>>> o2.carbon.ml.rest.api\src\main\java\org\wso2\carbon\ml\rest\api ) and >>>> can I access the dataset through the response's *"sourcePath" ?* >>>> >>>> Thank you. >>>> >>>> On Sun, Aug 7, 2016 at 8:06 PM, Supun Sethunga <sup...@wso2.com> wrote: >>>> >>>>> Hi Lakini, >>>>> >>>>> Please find all the APIs and their samples at [1]. Additionally, you >>>>> can always refer the existing UI to get an idea. for eg: you can check the >>>>> dataset page of the existing ML UI to see how the datasets are retrieved. >>>>> >>>>> [1] https://docs.wso2.com/display/ML110/REST+API+Guides >>>>> >>>>> Regards, >>>>> Supun. >>>>> >>>>> On Sat, Aug 6, 2016 at 11:06 AM, Lakini Senanayaka < >>>>> lakinisenanayak...@gmail.com> wrote: >>>>> >>>>>> Hi Upul, >>>>>> >>>>>> Could you please explain me the way to get the dataset from the >>>>>> carbon ML to the API which I am implementing. >>>>>> >>>>>> Thank you. >>>>>> >>>>>> On Thu, Aug 4, 2016 at 4:37 PM, Lakini Senanayaka < >>>>>> lakinisenanayak...@gmail.com> wrote: >>>>>> >>>>>>> Hi, >>>>>>> >>>>>>> This is the meeting minute for today's GSoC project review meeting. >>>>>>> >>>>>>> Upul reviewed the current progress of the project, "(Deep) Neural >>>>>>> Network Builder for WSO2 Machine Learner". >>>>>>> Below mentioned are the decisions made and the new tasks allocated >>>>>>> at the meeting. >>>>>>> >>>>>>> Assigned action items: >>>>>>> -Making the UI more user-friendly. >>>>>>> Ex: Physical lines of connections between layers should >>>>>>> follow the movements of the layers. >>>>>>> -Train the feed forward network for MNIST dataset. >>>>>>> -Testing the accuracy of the neural network model for different >>>>>>> optimization and updater algorithms. >>>>>>> -Handling Exceptions. >>>>>>> -Trying to show graphs related to the model, along with the >>>>>>> accuracy of the model in the output. >>>>>>> -Start writing documentation. >>>>>>> -Showing a demo for the ML team. >>>>>>> >>>>>>> Decisions made: >>>>>>> >>>>>>> -Higher priority should be given in implementing feed-forward >>>>>>> network than RNN. >>>>>>> -Upul will provide me necessary instructions on the way to >>>>>>> access the dataset which is inserted through the WSO2 ML console as NN >>>>>>> builder UI needs to access that dataset and use as the training dataset. >>>>>>> >>>>>>> Thank you. >>>>>>> >>>>>>> On Wed, Aug 3, 2016 at 5:59 AM, Lakini Senanayaka < >>>>>>> lakinisenanayak...@gmail.com> wrote: >>>>>>> >>>>>>>> Hi, >>>>>>>> >>>>>>>> Thank you very much for your kindness. >>>>>>>> >>>>>>>> I can schedule a meeting on this Thursday(4th-August-2016). >>>>>>>> >>>>>>>> Thank you. >>>>>>>> >>>>>>>> On Tue, Aug 2, 2016 at 10:39 AM, Upul Bandara <u...@wso2.com> >>>>>>>> wrote: >>>>>>>> >>>>>>>>> Hi, >>>>>>>>> >>>>>>>>> As you know we have only two weeks to complete GSOC projects. >>>>>>>>> Unfortunately, this project is a little bit lagging behind and it >>>>>>>>> looks >>>>>>>>> like you have to put some additional effort during the last two week. >>>>>>>>> In >>>>>>>>> order to support you, we though it is better to have a small F2F >>>>>>>>> meeting >>>>>>>>> with you and it will help you to sort out any issues currently you are >>>>>>>>> facing with. >>>>>>>>> >>>>>>>>> So can you please schedule a meeting with the ML team? >>>>>>>>> >>>>>>>>> Thanks, >>>>>>>>> Upul >>>>>>>>> >>>>>>>>> On Sat, Jul 30, 2016 at 5:45 PM, Nirmal Fernando <nir...@wso2.com> >>>>>>>>> wrote: >>>>>>>>> >>>>>>>>>> >>>>>>>>>> On Sat, Jul 30, 2016 at 1:43 PM, Lakini Senanayaka < >>>>>>>>>> lakinisenanayak...@gmail.com> wrote: >>>>>>>>>> >>>>>>>>>>> Hi, >>>>>>>>>>> >>>>>>>>>>> I have solved the above problem.I have referred the mail thread >>>>>>>>>>> *[Dev] >>>>>>>>>>> "Error 403 - Forbidden" when session expires in admin console*[1] in >>>>>>>>>>> dev mailing list.I couldn't upload jaggery files to the console >>>>>>>>>>> and when >>>>>>>>>>> I was trying to do it, it gave the* Error 403-Forbidden.* >>>>>>>>>>> >>>>>>>>>>> I have copied the jaggery files to >>>>>>>>>>> *<ML_HOME>/repository/deployment/server/jaggerapps* directory as >>>>>>>>>>> Supun instructed and it worked. >>>>>>>>>>> >>>>>>>>>>> Although I have implemented the back end for feed forward neural >>>>>>>>>>> network still it has some problems.It doesn't work well.Do I need >>>>>>>>>>> to add >>>>>>>>>>> the neural network algorithms to the >>>>>>>>>>> *org.wso2.carbon.ml.core.spark.algorithms* in Carbon ML ?Do I >>>>>>>>>>> need to follow the same pattern which is used by the Carbon ML when >>>>>>>>>>> implementing the back end coding and APIs? >>>>>>>>>>> >>>>>>>>>> >>>>>>>>>> Yes please. >>>>>>>>>> >>>>>>>>>>> >>>>>>>>>>> I will submit the demo ASAP. >>>>>>>>>>> >>>>>>>>>>> [1]https://mail.google.com/mail/u/0/#search/Re%3A+%5BDev%5D+ >>>>>>>>>>> %22Error+403++Forbidden%22+when+session+expires+in+admin%09c >>>>>>>>>>> onsole/155bfcbf7e98992f?projector=1 >>>>>>>>>>> >>>>>>>>>>> Thank you. >>>>>>>>>>> -- >>>>>>>>>>> KIND Regards, >>>>>>>>>>> *Lakini Senanayaka.* >>>>>>>>>>> >>>>>>>>>>> >>>>>>>>>> >>>>>>>>>> >>>>>>>>>> -- >>>>>>>>>> >>>>>>>>>> Thanks & regards, >>>>>>>>>> Nirmal >>>>>>>>>> >>>>>>>>>> Team Lead - WSO2 Machine Learner >>>>>>>>>> Associate Technical Lead - Data Technologies Team, WSO2 Inc. >>>>>>>>>> Mobile: +94715779733 >>>>>>>>>> Blog: http://nirmalfdo.blogspot.com/ >>>>>>>>>> >>>>>>>>>> >>>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> -- >>>>>>>>> Upul Bandara, >>>>>>>>> Associate Technical Lead, WSO2, Inc., >>>>>>>>> Mob: +94 715 468 345. >>>>>>>>> >>>>>>>> >>>>>>>> >>>>>>>> >>>>>>>> -- >>>>>>>> KIND Regards, >>>>>>>> *Lakini Senanayaka.* >>>>>>>> >>>>>>>> >>>>>>> >>>>>>> >>>>>>> -- >>>>>>> KIND Regards, >>>>>>> *Lakini Senanayaka.* >>>>>>> >>>>>>> >>>>>> >>>>>> >>>>>> -- >>>>>> KIND Regards, >>>>>> *Lakini Senanayaka.* >>>>>> >>>>>> >>>>> >>>>> >>>>> -- >>>>> *Supun Sethunga* >>>>> Senior Software Engineer >>>>> WSO2, Inc. >>>>> http://wso2.com/ >>>>> lean | enterprise | middleware >>>>> Mobile : +94 716546324 >>>>> Blog: http://supunsetunga.blogspot.com >>>>> >>>> >>>> >>>> >>>> -- >>>> KIND Regards, >>>> *Lakini Senanayaka.* >>>> >>>> >>> >>> >>> -- >>> Upul Bandara, >>> Associate Technical Lead, WSO2, Inc., >>> Mob: +94 715 468 345. >>> >> >> >> >> -- >> KIND Regards, >> *Lakini Senanayaka.* >> >> > > > -- > Upul Bandara, > Associate Technical Lead, WSO2, Inc., > Mob: +94 715 468 345. > -- KIND Regards, *Lakini Senanayaka.*
_______________________________________________ Dev mailing list Dev@wso2.org http://wso2.org/cgi-bin/mailman/listinfo/dev
