From awd at cs.cmu.edu Thu Jun 3 18:21:58 2010 From: awd at cs.cmu.edu (Artur Dubrawski) Date: Thu, 03 Jun 2010 18:21:58 -0400 Subject: [Research] Lab meeting: Wednesday June 8th Message-ID: <4C082B06.9030604@cs.cmu.edu> Speaker: Yi Zhang Title and Abstract: See below. Time: Noon, as usual Place: Karen W. will provide info Pizza: Yes See you all there! Artur --- Projection Penalties: Dimension Reduction without Loss Yi Zhang and Jeff Schneider Dimension reduction is popular for learning predictive models in high-dimensional spaces. It can highlight the relevant part of the feature space and avoid the curse of dimensionality. However, it can also be harmful because any reduction loses information. In this paper, we propose the \textit{projection penalty} framework to make use of dimension reduction without losing valuable information. Reducing the feature space before learning predictive models can be viewed as restricting the model search to some parameter subspace. The idea of projection penalties is that instead of restricting the search to a parameter subspace, we can search in the full space but penalize the projection distance to this subspace. Dimension reduction is used to guide the search, rather than to restrict it. We propose projection penalties for linear dimension reduction, and then generalize to kernel-based reduction and other nonlinear methods. We test projection penalties with various dimension reduction techniques in different prediction tasks, including principal component regression and partial least squares in regression tasks, kernel dimension reduction in face recognition, and latent topic modeling in text classification. Experimental results show that projection penalties are a more effective and reliable way to make use of dimension reduction techniques. From awd at cs.cmu.edu Thu Jun 3 18:24:41 2010 From: awd at cs.cmu.edu (Artur Dubrawski) Date: Thu, 03 Jun 2010 18:24:41 -0400 Subject: [Research] Lab meeting: Wednesday June *9*th In-Reply-To: <4C082B06.9030604@cs.cmu.edu> References: <4C082B06.9030604@cs.cmu.edu> Message-ID: <4C082BA9.4060905@cs.cmu.edu> Well, of course Wednesday is June 9th. Darned European calendars! A. Artur Dubrawski wrote: > Speaker: Yi Zhang > Title and Abstract: See below. > Time: Noon, as usual > Place: Karen W. will provide info > Pizza: Yes > > See you all there! > Artur > > --- > Projection Penalties: Dimension Reduction without Loss > Yi Zhang and Jeff Schneider > > Dimension reduction is popular for learning predictive models in > high-dimensional spaces. It can highlight the relevant part of the > feature space and avoid the curse of dimensionality. However, it can > also be harmful because any reduction loses information. In this paper, > we propose the \textit{projection penalty} framework to make use of > dimension reduction without losing valuable information. > > Reducing the feature space before learning predictive models can be > viewed as restricting the model search to some parameter subspace. The > idea of projection penalties is that instead of restricting the search > to a parameter subspace, we can search in the full space but penalize > the projection distance to this subspace. Dimension reduction is used to > guide the search, rather than to restrict it. > > We propose projection penalties for linear dimension reduction, and then > generalize to kernel-based reduction and other nonlinear methods. We > test projection penalties with various dimension reduction techniques in > different prediction tasks, including principal component regression and > partial least squares in regression tasks, kernel dimension reduction in > face recognition, and latent topic modeling in text classification. > Experimental results show that projection penalties are a more effective > and reliable way to make use of dimension reduction techniques. > _______________________________________________ > Research mailing list > Research at autonlab.org > https://www.autonlab.org/mailman/listinfo/research > > From krw at andrew.cmu.edu Fri Jun 4 08:30:56 2010 From: krw at andrew.cmu.edu (Karen Widmaier) Date: Fri, 4 Jun 2010 08:30:56 -0400 Subject: [Research] Lab meeting: Wednesday June *9*th In-Reply-To: <4C082BA9.4060905@cs.cmu.edu> References: <4C082B06.9030604@cs.cmu.edu> <4C082BA9.4060905@cs.cmu.edu> Message-ID: <001001cb03e1$c7bd2da0$573788e0$@cmu.edu> I've reserved NSH 1507 for Wednesday, June 9th from noon until 1:30pm. Karen -----Original Message----- From: research-bounces at autonlab.org [mailto:research-bounces at autonlab.org] On Behalf Of Artur Dubrawski Sent: Thursday, June 03, 2010 6:25 PM To: research at autonlab.org Subject: Re: [Research] Lab meeting: Wednesday June *9*th Well, of course Wednesday is June 9th. Darned European calendars! A. Artur Dubrawski wrote: > Speaker: Yi Zhang > Title and Abstract: See below. > Time: Noon, as usual > Place: Karen W. will provide info > Pizza: Yes > > See you all there! > Artur > > --- > Projection Penalties: Dimension Reduction without Loss > Yi Zhang and Jeff Schneider > > Dimension reduction is popular for learning predictive models in > high-dimensional spaces. It can highlight the relevant part of the > feature space and avoid the curse of dimensionality. However, it can > also be harmful because any reduction loses information. In this paper, > we propose the \textit{projection penalty} framework to make use of > dimension reduction without losing valuable information. > > Reducing the feature space before learning predictive models can be > viewed as restricting the model search to some parameter subspace. The > idea of projection penalties is that instead of restricting the search > to a parameter subspace, we can search in the full space but penalize > the projection distance to this subspace. Dimension reduction is used to > guide the search, rather than to restrict it. > > We propose projection penalties for linear dimension reduction, and then > generalize to kernel-based reduction and other nonlinear methods. We > test projection penalties with various dimension reduction techniques in > different prediction tasks, including principal component regression and > partial least squares in regression tasks, kernel dimension reduction in > face recognition, and latent topic modeling in text classification. > Experimental results show that projection penalties are a more effective > and reliable way to make use of dimension reduction techniques. > _______________________________________________ > Research mailing list > Research at autonlab.org > https://www.autonlab.org/mailman/listinfo/research > > _______________________________________________ Research mailing list Research at autonlab.org https://www.autonlab.org/mailman/listinfo/research From awd at cs.cmu.edu Thu Jun 10 11:42:37 2010 From: awd at cs.cmu.edu (Artur Dubrawski) Date: Thu, 10 Jun 2010 11:42:37 -0400 Subject: [Research] Auton Lab bike ride poll Message-ID: <4C1107ED.5020902@cs.cmu.edu> Hi, Please go to this site and let us know your preferences for the bike ride this weekend. http://doodle.com/77dtbnk3fggcyhqy Thanks Artur From awd at cs.cmu.edu Thu Jun 10 13:22:47 2010 From: awd at cs.cmu.edu (Artur Dubrawski) Date: Thu, 10 Jun 2010 13:22:47 -0400 Subject: [Research] Auton Lab bike ride poll In-Reply-To: <4C1107ED.5020902@cs.cmu.edu> References: <4C1107ED.5020902@cs.cmu.edu> Message-ID: <4C111F67.4080204@cs.cmu.edu> BTW, if you plan to bring people with you (e.g. significant others) please add the info on that using "Comments" feature of doodle. Thx. A. Artur Dubrawski wrote: > Hi, > > Please go to this site and let us know your preferences for the bike > ride this weekend. > > http://doodle.com/77dtbnk3fggcyhqy > > Thanks > Artur > > From awd at cs.cmu.edu Fri Jun 11 07:36:40 2010 From: awd at cs.cmu.edu (Artur Dubrawski) Date: Fri, 11 Jun 2010 07:36:40 -0400 Subject: [Research] Auton Lab bike ride poll In-Reply-To: <4C111F67.4080204@cs.cmu.edu> References: <4C1107ED.5020902@cs.cmu.edu> <4C111F67.4080204@cs.cmu.edu> Message-ID: <4C121FC8.9070506@cs.cmu.edu> It looks like we have enough interest to ride twice: on Friday as well as on Saturday! I am planning to do just that myself. For Friday, according to the doodle poll, we will not need any bike rentals. Let's then just meet at the level G2 of CIC garage at about 5:30pm. For Saturday, 4 people (TK, Thahir, Yi and Liang) will need rental bikes. There are at least 3 rental outlets along our route (see http://bikepittsburgh.com/bike-rental-locations) But, only one opens at 10am. I recommend that the 4 in need of a rental confirm the bike availability and perhaps make reservations (if they are taken - I am not sure of that) by phone today and arrange to go there by car to the Eliza Furnace ("jail trail") location. The rest of the group will assemble at 10am at G2 level of CIC garage and bike that way to get to the rental place location by or before 10:30am. Let me know if you'd have any questions or suggestions. See you! Artur Artur Dubrawski wrote: > BTW, if you plan to bring people with you (e.g. significant others) > please add the info on that using "Comments" feature of doodle. > > Thx. > A. > > Artur Dubrawski wrote: >> Hi, >> >> Please go to this site and let us know your preferences for the bike >> ride this weekend. >> >> http://doodle.com/77dtbnk3fggcyhqy >> >> Thanks >> Artur >> >> > > From emcfowla at andrew.cmu.edu Fri Jun 11 10:34:52 2010 From: emcfowla at andrew.cmu.edu (Edward McFowland III) Date: Fri, 11 Jun 2010 10:34:52 -0400 Subject: [Research] Auton Lab bike ride poll In-Reply-To: <4C121FC8.9070506@cs.cmu.edu> References: <4C1107ED.5020902@cs.cmu.edu> <4C111F67.4080204@cs.cmu.edu> <4C121FC8.9070506@cs.cmu.edu> Message-ID: <19D09BF1-D0E8-457F-9C05-145558D3955F@andrew.cmu.edu> Hey Artur, Can/Will we do this again? I just don't have the time this weekend and for that matter a few weeks. But I would like to be apart of this if we do it again. Ed On Jun 11, 2010, at 7:36 AM, Artur Dubrawski wrote: > It looks like we have enough interest to ride twice: on Friday as well > as on Saturday! > I am planning to do just that myself. > > For Friday, according to the doodle poll, we will not need any bike > rentals. > Let's then just meet at the level G2 of CIC garage at about 5:30pm. > > For Saturday, 4 people (TK, Thahir, Yi and Liang) will need rental > bikes. > There are at least 3 rental outlets along our route (see > http://bikepittsburgh.com/bike-rental-locations) > But, only one opens at 10am. I recommend that the 4 in need of a > rental > confirm the bike availability and perhaps make reservations (if they > are > taken - I am not sure of that) by phone today and arrange to go > there by > car to the Eliza Furnace ("jail trail") location. The rest of the > group > will > assemble at 10am at G2 level of CIC garage and bike that way to get > to the rental place location by or before 10:30am. > > Let me know if you'd have any questions or suggestions. > > See you! > Artur > > Artur Dubrawski wrote: >> BTW, if you plan to bring people with you (e.g. significant others) >> please add the info on that using "Comments" feature of doodle. >> >> Thx. >> A. >> >> Artur Dubrawski wrote: >>> Hi, >>> >>> Please go to this site and let us know your preferences for the bike >>> ride this weekend. >>> >>> http://doodle.com/77dtbnk3fggcyhqy >>> >>> Thanks >>> Artur >>> >>> >> >> > > _______________________________________________ > Research mailing list > Research at autonlab.org > https://www.autonlab.org/mailman/listinfo/research > From awd at cs.cmu.edu Fri Jun 11 10:38:41 2010 From: awd at cs.cmu.edu (Artur Dubrawski) Date: Fri, 11 Jun 2010 10:38:41 -0400 Subject: [Research] Auton Lab bike ride poll In-Reply-To: <19D09BF1-D0E8-457F-9C05-145558D3955F@andrew.cmu.edu> References: <4C1107ED.5020902@cs.cmu.edu> <4C111F67.4080204@cs.cmu.edu> <4C121FC8.9070506@cs.cmu.edu> <19D09BF1-D0E8-457F-9C05-145558D3955F@andrew.cmu.edu> Message-ID: <4C124A71.2030608@cs.cmu.edu> I hope people will like it tonight and tomorrow so there will be enough interest to repeat. Plus, we will probably have a lab picnic sometime late September. A. Edward McFowland III wrote: > Hey Artur, > > Can/Will we do this again? I just don't have the time this weekend > and for that matter a few weeks. But I would like to be apart of this > if we do it again. > > Ed > On Jun 11, 2010, at 7:36 AM, Artur Dubrawski wrote: > >> It looks like we have enough interest to ride twice: on Friday as well >> as on Saturday! >> I am planning to do just that myself. >> >> For Friday, according to the doodle poll, we will not need any bike >> rentals. >> Let's then just meet at the level G2 of CIC garage at about 5:30pm. >> >> For Saturday, 4 people (TK, Thahir, Yi and Liang) will need rental >> bikes. >> There are at least 3 rental outlets along our route (see >> http://bikepittsburgh.com/bike-rental-locations) >> But, only one opens at 10am. I recommend that the 4 in need of a rental >> confirm the bike availability and perhaps make reservations (if they are >> taken - I am not sure of that) by phone today and arrange to go there by >> car to the Eliza Furnace ("jail trail") location. The rest of the group >> will >> assemble at 10am at G2 level of CIC garage and bike that way to get >> to the rental place location by or before 10:30am. >> >> Let me know if you'd have any questions or suggestions. >> >> See you! >> Artur >> >> Artur Dubrawski wrote: >>> BTW, if you plan to bring people with you (e.g. significant others) >>> please add the info on that using "Comments" feature of doodle. >>> >>> Thx. >>> A. >>> >>> Artur Dubrawski wrote: >>>> Hi, >>>> >>>> Please go to this site and let us know your preferences for the bike >>>> ride this weekend. >>>> >>>> http://doodle.com/77dtbnk3fggcyhqy >>>> >>>> Thanks >>>> Artur >>>> >>>> >>> >>> >> >> _______________________________________________ >> Research mailing list >> Research at autonlab.org >> https://www.autonlab.org/mailman/listinfo/research >> > > From awd at cs.cmu.edu Mon Jun 14 12:36:17 2010 From: awd at cs.cmu.edu (Artur Dubrawski) Date: Mon, 14 Jun 2010 12:36:17 -0400 Subject: [Research] [Fwd: more ML students, faculty, alumni win ICML/COLT best paper awards] Message-ID: <4C165A81.5000409@cs.cmu.edu> ... and more than 45% of the awardees have deep Auton Lab roots! Way to go Purna, Sajid, Andrew, Drew and Geoff! Artur -------------- next part -------------- An embedded message was scrubbed... From: Tom Mitchell Subject: more ML students, faculty, alumni win ICML/COLT best paper awards Date: Mon, 14 Jun 2010 11:49:44 -0400 Size: 6319 URL: From sajidms at gmail.com Mon Jun 14 13:54:58 2010 From: sajidms at gmail.com (Sajid Siddiqi) Date: Mon, 14 Jun 2010 13:54:58 -0400 Subject: [Research] [Fwd: more ML students, faculty, alumni win ICML/COLT best paper awards] In-Reply-To: <4C165A81.5000409@cs.cmu.edu> References: <4C165A81.5000409@cs.cmu.edu> Message-ID: Thanks Artur :) It was a pleasant surprise indeed. Sajid On Mon, Jun 14, 2010 at 12:36 PM, Artur Dubrawski wrote: > ... and more than 45% of the awardees have deep Auton Lab roots! > > Way to go Purna, Sajid, Andrew, Drew and Geoff! > > Artur > > > > ---------- Forwarded message ---------- > From: Tom Mitchell > To: ml-all at cs.cmu.edu, scs-council at cs.cmu.edu > Date: Mon, 14 Jun 2010 11:49:44 -0400 > Subject: more ML students, faculty, alumni win ICML/COLT best paper awards > Dear Friends, > > Last week I sent a note congratulating Brian Ziebart on his paper > placing as runnerup for the 2010 ICML Best Student Paper Award. Since > then, I've learned the the ML department has done exceptionally well > this year in earning awards at next week's International Conference on > Machine Learning (ICML 2010), and the Computational Learning Theory > (COLT 2010) Conference. > > The list includes: > > Best Student Paper (COLT 2010) > Theoretical Justification of Popular Link Prediction Heuristics > Purnamrita Sarkar, Deepayan Chakrabarti, Andrew Moore > > Runnerup Best Student Paper (ICML 2010) > "Modeling Interaction via the Principle of Maximum Causal Entropy" > Brian Ziebart, Anind Dey, Drew Bagnell > > Best Paper (ICML 2010) > Hilbert Space Embeddings of Hidden Markov Models. > Le Song, Byron Boots, Sajid Siddiqi, Geoff Gordon, and Alex Smola > > > Heartiest congratulations to Purna, Deepay, Andrew, Brian, Anind, > Drew, Le Song, Byron, Sajid, and Geoff. ICML, COLT and NIPS are the > three top conferences in machine learning, and this is a great > achievement for these researchers. And of course it reflects very > positively on our department as a whole. > > Keep up the great work everybody! > > Tom > > -- > Tom M. Mitchell > Fredkin University Professor and Department Head > Machine Learning Department > Carnegie Mellon University > www.cs.cmu.edu/~tom > > > _______________________________________________ > Research mailing list > Research at autonlab.org > https://www.autonlab.org/mailman/listinfo/research > > -------------- next part -------------- An HTML attachment was scrubbed... URL: From dbagnell at ri.cmu.edu Mon Jun 14 16:24:38 2010 From: dbagnell at ri.cmu.edu (Drew Bagnell) Date: Mon, 14 Jun 2010 20:24:38 -0000 Subject: [Research] [Fwd: more ML students, faculty, alumni win ICML/COLT best paper awards] In-Reply-To: References: <4C165A81.5000409@cs.cmu.edu> Message-ID: <6AF782EA-9F46-4327-8022-F68958232881@ri.cmu.edu> On Jun 14, 2010, at 1:54 PM, Sajid Siddiqi wrote: > Thanks Artur :) It was a pleasant surprise indeed. > > Sajid Congrats to all-- great Auton news. > > On Mon, Jun 14, 2010 at 12:36 PM, Artur Dubrawski > wrote: > ... and more than 45% of the awardees have deep Auton Lab roots! Coincidence?! I think not. --d > > Way to go Purna, Sajid, Andrew, Drew and Geoff! > > Artur > > > > ---------- Forwarded message ---------- > From: Tom Mitchell > To: ml-all at cs.cmu.edu, scs-council at cs.cmu.edu > Date: Mon, 14 Jun 2010 11:49:44 -0400 > Subject: more ML students, faculty, alumni win ICML/COLT best paper > awards > Dear Friends, > > Last week I sent a note congratulating Brian Ziebart on his paper > placing as runnerup for the 2010 ICML Best Student Paper Award. Since > then, I've learned the the ML department has done exceptionally well > this year in earning awards at next week's International Conference on > Machine Learning (ICML 2010), and the Computational Learning Theory > (COLT 2010) Conference. > > The list includes: > > Best Student Paper (COLT 2010) > Theoretical Justification of Popular Link Prediction Heuristics > Purnamrita Sarkar, Deepayan Chakrabarti, Andrew Moore > > Runnerup Best Student Paper (ICML 2010) > "Modeling Interaction via the Principle of Maximum Causal Entropy" > Brian Ziebart, Anind Dey, Drew Bagnell > > Best Paper (ICML 2010) > Hilbert Space Embeddings of Hidden Markov Models. > Le Song, Byron Boots, Sajid Siddiqi, Geoff Gordon, and Alex Smola > > > Heartiest congratulations to Purna, Deepay, Andrew, Brian, Anind, > Drew, Le Song, Byron, Sajid, and Geoff. ICML, COLT and NIPS are the > three top conferences in machine learning, and this is a great > achievement for these researchers. And of course it reflects very > positively on our department as a whole. > > Keep up the great work everybody! > > Tom > > -- > Tom M. Mitchell > Fredkin University Professor and Department Head > Machine Learning Department > Carnegie Mellon University > www.cs.cmu.edu/~tom > > > _______________________________________________ > Research mailing list > Research at autonlab.org > https://www.autonlab.org/mailman/listinfo/research > > > _______________________________________________ > Research mailing list > Research at autonlab.org > https://www.autonlab.org/mailman/listinfo/research -------------- next part -------------- An HTML attachment was scrubbed... URL: