tag:blogger.com,1999:blog-19803222.post9119507162108309109..comments2024-03-18T01:45:45.724-06:00Comments on natural language processing blog: Convex or not, my schizophrenic personalityhalhttp://www.blogger.com/profile/02162908373916390369noreply@blogger.comBlogger11125tag:blogger.com,1999:blog-19803222.post-32251115353991207382010-01-10T00:44:54.298-07:002010-01-10T00:44:54.298-07:00I found your blog on google and read a few of your...I found your blog on google and read a few of your other posts. I just added you to my Google News Reader. Keep up the good work. Look forward to reading more from you in the future. Please come visit my site <a href="http://www.lexingtonfayettedirectory.com" rel="nofollow">Louisville Phone Book</a> when yourr8004https://www.blogger.com/profile/17668784480317088693noreply@blogger.comtag:blogger.com,1999:blog-19803222.post-4787774614535557102010-01-10T00:44:39.015-07:002010-01-10T00:44:39.015-07:00I found your blog on google and read a few of your...I found your blog on google and read a few of your other posts. I just added you to my Google News Reader. Keep up the good work. Look forward to reading more from you in the future. Please come visit my site <a href="http://www.lexingtonfayettedirectory.com" rel="nofollow">Kentuchy KY Phone Directory</a> when yourr8004https://www.blogger.com/profile/17668784480317088693noreply@blogger.comtag:blogger.com,1999:blog-19803222.post-11641277892197895352009-11-27T01:59:25.249-07:002009-11-27T01:59:25.249-07:00You do have a point here :) I admire the stuff you...You do have a point here :) I admire the stuff you post and the quality information you offer in your blog! Keep up the good work dude. Please come visit my site <a href="http://www.guidesacramento.com" rel="nofollow">Fresno Web Link </a> when you got time.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-19803222.post-76465305732796994112009-11-27T01:59:10.978-07:002009-11-27T01:59:10.978-07:00You do have a point here :) I admire the stuff you...You do have a point here :) I admire the stuff you post and the quality information you offer in your blog! Keep up the good work dude. Please come visit my site <a href="http://www.guidesacramento.com" rel="nofollow">Fresno Business Search Engine</a> when you got time.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-19803222.post-50132980752177563582009-11-17T04:13:10.519-07:002009-11-17T04:13:10.519-07:00thnxxx for sharing the information
public speakin...thnxxx for sharing the information<br /><br /><a href="http://anuragaggarwal.com/" rel="nofollow">public speaking</a>Adminhttps://www.blogger.com/profile/17529312110752417139noreply@blogger.comtag:blogger.com,1999:blog-19803222.post-88470310249161454902009-11-14T02:25:20.979-07:002009-11-14T02:25:20.979-07:00Many institutions limit access to their online inf...Many institutions limit access to their online information. Making this information available will be an asset to all.Research Papershttp://researchpaperspot.com/noreply@blogger.comtag:blogger.com,1999:blog-19803222.post-43246236480960401042009-11-11T17:00:00.189-07:002009-11-11T17:00:00.189-07:00Convexity still matters, but we care less about it...Convexity still matters, but we care less about it when doing things that are never going to be convex. For supervised learning, we can get nice convex optimization for basically anything we want, and so we are happy. Why deal with non-convexity?<br /><br />For the more complex graphical models, could we ever hope to achieve convexity? No. If we could define a convex model, then we would. But since we couldn't learn these models without using non-convex objectives, we are willing to settle for non-convexity.<br /><br />The interesting tension comes when you have reasonable convex approximations for non-convex objectives. There are a number of papers that need to make this decision: solve the non-convex objective, or solve exactly a convex approximation.<br /><br />As we continue to look at these complex graphical models, the understanding of issues with non-convexity will become increasingly important.Mark Dredzehttps://www.blogger.com/profile/03294614528648962173noreply@blogger.comtag:blogger.com,1999:blog-19803222.post-65889778412338727012009-10-30T22:11:41.811-06:002009-10-30T22:11:41.811-06:00Wow! Thank you! I always wanted to write in my sit...Wow! Thank you! I always wanted to write in my site something like that. Can I take part of your post to my blog? Please come visit my site <a href="http://www.businessesindianapolis.com" rel="nofollow">Jacksonville Business Directory Forum Blog Classifieds</a> when you got time.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-19803222.post-43378653302549789052009-10-24T00:58:49.297-06:002009-10-24T00:58:49.297-06:00love your blog! Very cool! Please come visit my si...love your blog! Very cool! Please come visit my site <a href="http://www.usaustintexas.com" rel="nofollow">Austin Business Directory</a> when you got time.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-19803222.post-12448568789687478792009-10-24T00:58:18.308-06:002009-10-24T00:58:18.308-06:00love your blog! Very cool! Please come visit my si...love your blog! Very cool! Please come visit my site <a href="http://www.usaustintexas.com" rel="nofollow">Business Trade Guide of Columbus Ohio OH</a> when you got time.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-19803222.post-88784983973502154502009-10-22T10:32:05.365-06:002009-10-22T10:32:05.365-06:00Maybe I can answer the question at the end about L...Maybe I can answer the question at the end about LDA. Because of the way variational inference for LDA is formulated as an optimization problem, you need to initialize with something which breaks the symmetry and how you do this matters which is why AFAIK Blei's lda-c distribution by default still uses the seeding method you mentioned.<br /><br />A lot of people have lately moved towards using collapsed Gibbs sampling. Initialization seems to matter less here, and you needn't break any symmetries since the non-determinism will take care of that for you.Anonymousnoreply@blogger.com