mahout-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Rajesh Nikam <rajeshni...@gmail.com>
Subject Re: SGD: Logistic regression package in Mahout
Date Fri, 19 Oct 2012 11:47:21 GMT
Hi Ted,

Please update once SGD parsing issue is fixed.

Thanks
Rajesh

On Wed, Oct 17, 2012 at 2:22 PM, Rajesh Nikam <rajeshnikam@gmail.com> wrote:

> Hello Ted,
>
> Thanks for investigating into it.
> I would look forward for further analysis and fix in SGD.
>
> I appreciate your efforts in looking into it.
>
> Thanks,
> Rajesh
>
>
>
> On Tue, Oct 16, 2012 at 10:23 PM, Ted Dunning <ted.dunning@gmail.com>wrote:
>
>> Rajesh,
>>
>> In the testing that I did, I ran 100, 1000 and 10,000 passes through the
>> data.  All produced identical results.  Thus it isn't an issue of SGD
>> converging.
>>
>> I also did a parameter scan of lambda and saw no effect.
>>
>> I also did the standard thing in R with glm and got the expected (correct)
>> results.
>>
>> I haven't looked yet in detail, but I really suspect that the reading of
>> the data is horked.  This is exactly how that behaves.
>>
>> On Tue, Oct 16, 2012 at 4:49 AM, Rajesh Nikam <rajeshnikam@gmail.com>
>> wrote:
>>
>> > Hi Ted,
>> >
>> > I was thinking, this might be due to having only 100 instances for
>> > training.
>> >
>> > So I have created test set with two classes having ~49K instances,
>> included
>> > all features as predictors.
>> > PFA sgd.grps.zip with test file.
>> >
>> > mahout trainlogistic --input /usr/local/mahout/trainme/sgd-grps.csv
>> > --output /usr/local/mahout/trainme/sgd-grps.model --target class
>> > --categories 2 --features 128 --types n --predictors a1 a2 a3 a4 a5 a6
>> a7
>> > a8 a9 a10 a11 a12 a13 a14 a15 a16 a17 a18 a19 a20 a21 a22 a23 a24 a25
>> a26
>> > a27 a28 a29 a30 a31 a32 a33 a34 a35 a36 a37 a38 a39 a40 a41 a42 a43 a44
>> a45
>> > a46 a47 a48 a49 a50 a51 a52 a53 a54 a55 a56 a57 a58 a59 a60 a61 a62 a63
>> a64
>> > a65 a66 a67 a68 a69 a70 a71 a72 a73 a74 a75 a76 a77 a78 a79 a80 a81 a82
>> a83
>> > a84 a85 a86 a87 a88 a89 a90 a91 a92 a93 a94 a95 a96 a97 a98 a99 a100
>> a101
>> > a102 a103 a104 a105 a106 a107 a108 a109 a110 a111 a112 a113 a114 a115
>> a116
>> > a117 a118 a119 a120 a121 a122 a123 a124 a125 a126 a127
>> >
>> >
>> > mahout runlogistic --input /usr/local/mahout/trainme/sgd-grps.csv
>> --model
>> > /usr/local/mahout/trainme/sgd-grps.model --auc --confusion
>> >
>> > Still the results are similar, it classifies everything as class_1.
>> >
>> > AUC = 0.50
>> > confusion: [[*26563.0, 23006.0*], [0.0, 0.0]]
>> > entropy: [[-0.0, -0.0], [-46.1, -21.4]]
>> >
>> > I am not sure why this is failing all the time.
>> >
>> > Looking forward for your reply.
>> >
>> > Thanks
>> > Rajesh
>> >
>> >
>> >
>> > On Tue, Oct 16, 2012 at 3:57 AM, Ted Dunning <ted.dunning@gmail.com>
>> > wrote:
>> >
>> > > I would love to help and will before long.  Just can't do it in the
>> first
>> > > part of this week.
>> > >
>> > > On Mon, Oct 15, 2012 at 6:28 AM, Rajesh Nikam <rajeshnikam@gmail.com>
>> > > wrote:
>> > >
>> > > > Hello,
>> > > >
>> > > > I have asked below question on issue with using sgd on mahout forum.
>> > > >
>> > > > Similar issue with sgd is reported by
>> > > >
>> > > >
>> > >
>> >
>> http://stackoverflow.com/questions/11221436/using-sgd-classifier-in-mahout
>> > > >
>> > > > Even below link has similar output:
>> > > >
>> > > > AUC = 0.57*confusion: [[27.0, 13.0], [0.0, 0.0]]*
>> > > > entropy: [[-0.4, -0.3], [-1.2, -0.7]]
>> > > >
>> > > >
>> > > >
>> > http://sujitpal.blogspot.in/2012/09/learning-mahout-classification.html
>> > > >
>> > > > I am still wannder confusion how then this model works and used by
>> > many ?
>> > > > Not able to get any points on how to use SGD that generates
>> effective
>> > > > model.
>> > > >
>> > > > Could someone point out what is missing in input file or provided
>> > > > parameters.
>> > > >
>> > > > I appreciate your help.
>> > > >
>> > > > Below is description of steps that I followed.
>> > > >
>> > > > PF Attached uses input files for experiment.
>> > > >
>> > > > I am using Iris Plants Database from Michael Marshall. PFA
>> iris.arff.
>> > > > Converted this to csv file just by updating header:
>> iris-3-classes.csv
>> > > >
>> > > > mahout org.apache.mahout.classifier.
>> > > > sgd.TrainLogistic --input
>> > > /usr/local/mahout/trunk/*iris-3-classes.csv*--features 4 --output
>> > > /usr/local/mahout/trunk/
>> > > > *iris-3-classes.model* --target class *--categories 3* --predictors
>> > > > sepallength sepalwidth petallength petalwidth --types n
>> > > >
>> > > > >> it gave following error.
>> > > > Exception in thread "main" java.lang.IllegalArgumentException: Can
>> only
>> > > > call classifyScalar with two categories
>> > > >
>> > > > Now created csv with only 2 classes. PFA iris-2-classes.csv
>> > > >
>> > > > >> trained iris-2-classes.csv with sgd
>> > > >
>> > > > mahout org.apache.mahout.classifier.sgd.TrainLogistic --input
>> > > > /usr/local/mahout/trunk/*iris-2-classes.csv* --features 4 --output
>> > > > /usr/local/mahout/trunk/*iris-2-classes.mode*l --target class
>> > > *--categories
>> > > > 2* --predictors sepallength sepalwidth petallength petalwidth
>> --types n
>> > > >
>> > > > mahout runlogistic --input
>> /usr/local/mahout/trunk/iris-2-classes.csv
>> > > > --model /usr/local/mahout/trunk/iris-2-classes.model --auc
>> --confusion
>> > > >
>> > > > AUC = 0.14
>> > > > confusion: [[50.0, 50.0], [0.0, 0.0]]
>> > > > entropy: [[-0.6, -0.3], [-0.8, -0.4]]
>> > > >
>> > > > >> AUC seems to poor. Now changed --predictors
>> > > >
>> > > > mahout org.apache.mahout.classifier.sgd.TrainLogistic --input
>> > > > /usr/local/mahout/trunk/*iris-2-classes.csv* --features 4 --output
>> > > > /usr/local/mahout/trunk/*iris-2-classes.mode*l --target class
>> > > *--categories
>> > > > 2* --predictors sepalwidth petallength --types n
>> > > >
>> > > > mahout runlogistic --input
>> /usr/local/mahout/trunk/iris-2-classes.csv
>> > > > --model /usr/local/mahout/trunk/iris-2-classes.model --auc
>> --confusion
>> > > > --scores
>> > > >
>> > > > AUC = 0.80
>> > > > *confusion: [[50.0, 50.0], [0.0, 0.0]]*
>> > > > entropy: [[-0.7, -0.3], [-0.7, -0.4]]
>> > > >
>> > > > This model classifies everything as category 1 which of no use.
>> > > >
>> > > > Thanks
>> > > > Rajesh
>> > > >
>> > > >
>> > > >
>> > > >
>> > >
>> >
>>
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message