tag:blogger.com,1999:blog-37324607.post4563278228395975570..comments2023-11-03T08:31:23.698-04:00Comments on Data Mining in MATLAB: L-1 Linear RegressionWill Dwinnellhttp://www.blogger.com/profile/03379859054257561952noreply@blogger.comBlogger9125tag:blogger.com,1999:blog-37324607.post-54218071964821639502016-03-16T12:17:31.091-04:002016-03-16T12:17:31.091-04:00Hello,
I really much want to know how does those f...Hello,<br />I really much want to know how does those function works with 3D data points??Song Snhttps://www.blogger.com/profile/07360945361359740166noreply@blogger.comtag:blogger.com,1999:blog-37324607.post-39327703225636356412016-03-11T09:36:18.033-05:002016-03-11T09:36:18.033-05:00This comment has been removed by the author.Song Snhttps://www.blogger.com/profile/07360945361359740166noreply@blogger.comtag:blogger.com,1999:blog-37324607.post-77581779852819471682012-05-22T12:13:17.889-04:002012-05-22T12:13:17.889-04:00Okay, I've found several iterative algorithms,...Okay, I've found several iterative algorithms, for example in the survey by Dasgupta et al. <a rel="nofollow">http://mpra.ub.uni-muenchen.de/1781/1/MPRA_paper_1781.pdf</a>. The 2004 paper by Li and Arse <a rel="nofollow">http://downloads.hindawi.com/journals/asp/2004/948982.pdf</a> points out that the naive algorithm can get stuck in a local minimum and suggests improvements. Several methods are compared by Kuzmanovic et al. <a rel="nofollow">http://journal.aplimat.com/volume_2_2009/.../Kuzmanovic_Sabo.pdf</a>Jim Van Zandthttps://www.blogger.com/profile/01724923409800351670noreply@blogger.comtag:blogger.com,1999:blog-37324607.post-25065236982215901002012-05-21T13:06:55.608-04:002012-05-21T13:06:55.608-04:00Your L1LinearRegression function finds the L1 solu...Your L1LinearRegression function finds the L1 solution using a sequence of weighted least squares problems. The more standard approach is to solve a linear programming problem. Could you compare the two methods? E.g. is your method guaranteed to converge? How do run times compare?Jim Van Zandthttps://www.blogger.com/profile/01724923409800351670noreply@blogger.comtag:blogger.com,1999:blog-37324607.post-49866567357033928462010-01-17T04:11:07.920-05:002010-01-17T04:11:07.920-05:00Hi All,
I was wondering how do I get such informa...Hi All,<br /><br />I was wondering how do I get such information that the least squares method is the maximum likelihood estimation if the errors are laplace... am startin in this field n I like to understand what am doing.. not just plug data into equations...<br /><br />Thanks <br />E.A.Unknownhttps://www.blogger.com/profile/02250453852802857545noreply@blogger.comtag:blogger.com,1999:blog-37324607.post-87315842808414919142008-04-16T16:44:00.000-04:002008-04-16T16:44:00.000-04:00Thanks, Finn Årup Nielsen!Thanks, Finn Årup Nielsen!Will Dwinnellhttps://www.blogger.com/profile/03379859054257561952noreply@blogger.comtag:blogger.com,1999:blog-37324607.post-41559117432111345632008-04-14T11:53:00.000-04:002008-04-14T11:53:00.000-04:00The L-1 linear regression solution is the maximum ...The L-1 linear regression solution is the maximum likelihood solution when the errors are Laplace (double exponential) distributed.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-37324607.post-40512067906081112462007-11-27T18:16:00.000-05:002007-11-27T18:16:00.000-05:00I believe (someone correct me if I'm wrong) that t...I believe (someone correct me if I'm wrong) that the least squares regression is only the maximum likelihood solution when the errors have a Gaussian distribution.<BR/><BR/>Regardless, the performance function should be defined by the needs of the analysis, and the least absolute error is a logical choice for many problems.Will Dwinnellhttps://www.blogger.com/profile/03379859054257561952noreply@blogger.comtag:blogger.com,1999:blog-37324607.post-81786768700529918002007-11-11T03:35:00.000-05:002007-11-11T03:35:00.000-05:00it is indeed true that least squares regression re...it is indeed true that least squares regression results in a particularly easy computation of the optimal weights of the hyperplane to be learned. However, one should not forget that the least squares solution to regression is also the maximum likelihood solution!Anonymousnoreply@blogger.com