Re: [Scikit-learn-general] What do SGDClassifier weights do mathematically?
Sounds good. I have to think about the answers. I'm trying to find out whether replicating each Xi as pi rows with positive class and ni rows with negative class corresponds to any unreplicated version but with sample_weights. Can you confirm and what weights would pi/ni correspond to? Am 25.06.2015 11:43 schrieb Joel Nothman joel.noth...@gmail.com: Across models, weights should be implemented such that duplicating samples would give identical results to corresponding integer weights. That is true here, to my understanding, if we remove the stochasticity such that all identical samples have their update occur at once. On 25 June 2015 at 19:28, Daniel Sullivan dbsulliva...@gmail.com wrote: Hi Anton, The update for each sample is just multiplied by the sample_weight and the class_weight before it's applied to the weight vector (coef_). So if your sample is [1, 2, 3], your gradient is .1, your weight for this sample is .2, and your eta is 5.0 your weight_vector (coef_) would be updated by [+(1 * .1 * .2 * 5.0),+(2 * .1 * .2 * 5.0),+(3 *.1 * .2 * 5.0)]. Does that help? Danny On Thu, Jun 25, 2015 at 10:19 AM, Anton Suchaneck a.suchan...@gmail.com wrote: Hello! How can I find out what the precise mathematical treatment of the sample_weights for SGDClassifier in the partial_fit setting is? Cheers, Anton -- Monitor 25 network devices or servers for free with OpManager! OpManager is web-based network management software that monitors network devices and physical virtual servers, alerts via email sms for fault. Monitor 25 devices for free with no restriction. Download now http://ad.doubleclick.net/ddm/clk/292181274;119417398;o ___ Scikit-learn-general mailing list Scikit-learn-general@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/scikit-learn-general -- Monitor 25 network devices or servers for free with OpManager! OpManager is web-based network management software that monitors network devices and physical virtual servers, alerts via email sms for fault. Monitor 25 devices for free with no restriction. Download now http://ad.doubleclick.net/ddm/clk/292181274;119417398;o ___ Scikit-learn-general mailing list Scikit-learn-general@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/scikit-learn-general -- Monitor 25 network devices or servers for free with OpManager! OpManager is web-based network management software that monitors network devices and physical virtual servers, alerts via email sms for fault. Monitor 25 devices for free with no restriction. Download now http://ad.doubleclick.net/ddm/clk/292181274;119417398;o ___ Scikit-learn-general mailing list Scikit-learn-general@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/scikit-learn-general -- Monitor 25 network devices or servers for free with OpManager! OpManager is web-based network management software that monitors network devices and physical virtual servers, alerts via email sms for fault. Monitor 25 devices for free with no restriction. Download now http://ad.doubleclick.net/ddm/clk/292181274;119417398;o___ Scikit-learn-general mailing list Scikit-learn-general@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
[Scikit-learn-general] What do SGDClassifier weights do mathematically?
Hello! How can I find out what the precise mathematical treatment of the sample_weights for SGDClassifier in the partial_fit setting is? Cheers, Anton -- Monitor 25 network devices or servers for free with OpManager! OpManager is web-based network management software that monitors network devices and physical virtual servers, alerts via email sms for fault. Monitor 25 devices for free with no restriction. Download now http://ad.doubleclick.net/ddm/clk/292181274;119417398;o___ Scikit-learn-general mailing list Scikit-learn-general@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
Re: [Scikit-learn-general] What do SGDClassifier weights do mathematically?
Hi Anton, The update for each sample is just multiplied by the sample_weight and the class_weight before it's applied to the weight vector (coef_). So if your sample is [1, 2, 3], your gradient is .1, your weight for this sample is .2, and your eta is 5.0 your weight_vector (coef_) would be updated by [+(1 * .1 * .2 * 5.0),+(2 * .1 * .2 * 5.0),+(3 *.1 * .2 * 5.0)]. Does that help? Danny On Thu, Jun 25, 2015 at 10:19 AM, Anton Suchaneck a.suchan...@gmail.com wrote: Hello! How can I find out what the precise mathematical treatment of the sample_weights for SGDClassifier in the partial_fit setting is? Cheers, Anton -- Monitor 25 network devices or servers for free with OpManager! OpManager is web-based network management software that monitors network devices and physical virtual servers, alerts via email sms for fault. Monitor 25 devices for free with no restriction. Download now http://ad.doubleclick.net/ddm/clk/292181274;119417398;o ___ Scikit-learn-general mailing list Scikit-learn-general@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/scikit-learn-general -- Monitor 25 network devices or servers for free with OpManager! OpManager is web-based network management software that monitors network devices and physical virtual servers, alerts via email sms for fault. Monitor 25 devices for free with no restriction. Download now http://ad.doubleclick.net/ddm/clk/292181274;119417398;o ___ Scikit-learn-general mailing list Scikit-learn-general@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
Re: [Scikit-learn-general] What do SGDClassifier weights do mathematically?
Across models, weights should be implemented such that duplicating samples would give identical results to corresponding integer weights. That is true here, to my understanding, if we remove the stochasticity such that all identical samples have their update occur at once. On 25 June 2015 at 19:28, Daniel Sullivan dbsulliva...@gmail.com wrote: Hi Anton, The update for each sample is just multiplied by the sample_weight and the class_weight before it's applied to the weight vector (coef_). So if your sample is [1, 2, 3], your gradient is .1, your weight for this sample is .2, and your eta is 5.0 your weight_vector (coef_) would be updated by [+(1 * .1 * .2 * 5.0),+(2 * .1 * .2 * 5.0),+(3 *.1 * .2 * 5.0)]. Does that help? Danny On Thu, Jun 25, 2015 at 10:19 AM, Anton Suchaneck a.suchan...@gmail.com wrote: Hello! How can I find out what the precise mathematical treatment of the sample_weights for SGDClassifier in the partial_fit setting is? Cheers, Anton -- Monitor 25 network devices or servers for free with OpManager! OpManager is web-based network management software that monitors network devices and physical virtual servers, alerts via email sms for fault. Monitor 25 devices for free with no restriction. Download now http://ad.doubleclick.net/ddm/clk/292181274;119417398;o ___ Scikit-learn-general mailing list Scikit-learn-general@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/scikit-learn-general -- Monitor 25 network devices or servers for free with OpManager! OpManager is web-based network management software that monitors network devices and physical virtual servers, alerts via email sms for fault. Monitor 25 devices for free with no restriction. Download now http://ad.doubleclick.net/ddm/clk/292181274;119417398;o ___ Scikit-learn-general mailing list Scikit-learn-general@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/scikit-learn-general -- Monitor 25 network devices or servers for free with OpManager! OpManager is web-based network management software that monitors network devices and physical virtual servers, alerts via email sms for fault. Monitor 25 devices for free with no restriction. Download now http://ad.doubleclick.net/ddm/clk/292181274;119417398;o___ Scikit-learn-general mailing list Scikit-learn-general@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/scikit-learn-general