From mjbaysek at cs.cmu.edu Wed Sep 9 10:56:36 2009 From: mjbaysek at cs.cmu.edu (Michael J. Baysek) Date: Wed, 09 Sep 2009 10:56:36 -0400 Subject: [Research] Auton Lab Picnic Message-ID: <4AA7C224.7080601@cs.cmu.edu> Hi Again Lab, We're planning to do the lab picnic next week at North Park. North Park is a very cool place with lots of things to do. We plan to have lunch together, and then probably break into groups who want to do different things. If you are interested in any of the activities below, reply (or reply all) to this email. Possible dates, times for the picnic, somewhat flexible: Fri Sep 18: 12 PM-4 PM Sat Sep 19: 10 AM-2 PM Please let me know which times do, or do not work for you so we can pick a day and time that is best compatible for everyone. -Mike ------------------------------------------- Peripheral Info [shamelessly pasted from the last email]: WHAT: Here is a long but partial list of amenities. * Food, of course! * Bike trails (bike rental available) * Fishing * Tennis * Nature Walks * Driving ranges * Batting cages * Kayaking (kayak, canoe, rowboat and more are available for rent) * Swimming Pool * Horseback riding * Probably board games, and other games for whoever wants to stick close to the picnic grove. Have any ideas? Let us know! WHERE: North Park: http://www.alleghenycounty.us/parks/npfac.aspx North Park is the largest park in Allegheny County, covering over 3000 acres, and complete with a 72 acre lake, bike trails, sports fields, picnic groves and more! We will have a ride leaving from Oakland if you need one. Please reply as soon as possible so we can start to get a head count, and figure out which date is going to work for everybody. We will keep you posted of any changes or updates. And, bring your family if you can - There will be fun for everyone :) From sabhnani+ at cs.cmu.edu Fri Sep 11 14:09:58 2009 From: sabhnani+ at cs.cmu.edu (Robin Sabhnani) Date: Fri, 11 Sep 2009 14:09:58 -0400 Subject: [Research] optimization problem.. In-Reply-To: <4AA7ED19.50402@cs.cmu.edu> References: <4AA0111A.8010907@cs.cmu.edu> <4AA040C7.8060106@cs.cmu.edu> <4AA169CC.3020507@cs.cmu.edu> <4AA7ED19.50402@cs.cmu.edu> Message-ID: <4AAA9276.7040104@cs.cmu.edu> Given a 2-D matrix containing integers, find the submatrix S (formed by a subset of rows and columns) that maximizes the sum of S. Any ideas .. Thanks in advance. Robin From neill at cs.cmu.edu Fri Sep 11 16:23:39 2009 From: neill at cs.cmu.edu (Daniel B. Neill) Date: Fri, 11 Sep 2009 16:23:39 -0400 (EDT) Subject: [Research] optimization problem.. In-Reply-To: <4AAA9276.7040104@cs.cmu.edu> References: <4AA0111A.8010907@cs.cmu.edu> <4AA040C7.8060106@cs.cmu.edu> <4AA169CC.3020507@cs.cmu.edu> <4AA7ED19.50402@cs.cmu.edu> <4AAA9276.7040104@cs.cmu.edu> Message-ID: I think this problem has been solved by Agarwal et al: http://www.cs.duke.edu/~jeffp/papers/stat-disc-SODA06.pdf Best, Daniel On Fri, 11 Sep 2009, Robin Sabhnani wrote: > Given a 2-D matrix containing integers, find the submatrix S (formed by > a subset of rows and columns) that maximizes the sum of S. > > Any ideas .. > > Thanks in advance. > Robin > _______________________________________________ > Research mailing list > Research at autonlab.org > https://www.autonlab.org/mailman/listinfo/research > > From mjbaysek at cs.cmu.edu Mon Sep 14 11:28:44 2009 From: mjbaysek at cs.cmu.edu (Michael J. Baysek) Date: Mon, 14 Sep 2009 11:28:44 -0400 Subject: [Research] Auton Lab Picnic is THIS WEEK In-Reply-To: <4AA7C224.7080601@cs.cmu.edu> References: <4AA7C224.7080601@cs.cmu.edu> Message-ID: <4AAE612C.405@cs.cmu.edu> Hi Lab, If you are receiving this email you are invited to the lab picnic. It's going to be a great time of food and fun! So far we received very few responses for your preferences. If you haven't already please email me today by 4PM, and let me know which days you are available, and what activities you might be interested in. Read the original mail below, for all the details. Mike Michael J. Baysek wrote, On 09/09/2009 10:56 AM: > Hi Again Lab, > > > We're planning to do the lab picnic next week at North Park. North Park > is a very cool place with lots of things to do. We plan to have lunch > together, and then probably break into groups who want to do different > things. If you are interested in any of the activities below, reply (or > reply all) to this email. > > > Possible dates, times for the picnic, somewhat flexible: > > > Fri Sep 18: 12 PM-4 PM > Sat Sep 19: 10 AM-2 PM > > > Please let me know which times do, or do not work for you so we can pick > a day and time that is best compatible for everyone. > > -Mike > > > ------------------------------------------- > > Peripheral Info [shamelessly pasted from the last email]: > > WHAT: Here is a long but partial list of amenities. > > > * Food, of course! > * Bike trails (bike rental available) > * Fishing > * Tennis > * Nature Walks > * Driving ranges > * Batting cages > * Kayaking (kayak, canoe, rowboat and more are available for rent) > * Swimming Pool > * Horseback riding > * Probably board games, and other games for whoever wants to stick > close to the picnic grove. Have any ideas? Let us know! > > > WHERE: North Park: http://www.alleghenycounty.us/parks/npfac.aspx > > > North Park is the largest park in Allegheny County, covering over 3000 > acres, > and complete with a 72 acre lake, bike trails, sports fields, picnic groves > and more! We will have a ride leaving from Oakland if you need one. > > > Please reply as soon as possible so we can start to get a head count, and > figure out which date is going to work for everybody. We will keep you > posted of any changes or updates. > > > And, bring your family if you can - There will be fun for everyone :) > > _______________________________________________ > Research mailing list > Research at autonlab.org > https://www.autonlab.org/mailman/listinfo/research > > From schneide at cs.cmu.edu Tue Sep 15 14:08:14 2009 From: schneide at cs.cmu.edu (Jeff Schneider) Date: Tue, 15 Sep 2009 14:08:14 -0400 Subject: [Research] GP and all of statistics books Message-ID: <4AAFD80E.9060308@cs.cmu.edu> Sorry for the spam, but I'm looking for two of my books and I'm not sure who I loaned them too. If you have either of these would you return them to me: Gaussian Processes for Machine Learning C. Rasmussen and C. Williams MIT Press, 2006. All of Statistics L. Wasserman Springer, 2004. thanks, Jeff. From mjbaysek at cs.cmu.edu Tue Sep 15 14:17:47 2009 From: mjbaysek at cs.cmu.edu (Michael J. Baysek) Date: Tue, 15 Sep 2009 14:17:47 -0400 Subject: [Research] Auton Lab Picnic is SATURDAY 10 AM to 2 PM Message-ID: <4AAFDA4B.7070609@cs.cmu.edu> Hi Lab, The picnic will be on Saturday, but the location is changed to MORAINE STATE PARK. There is also lots to do there, so check the website. If you want to organize any special adventures, let me know! At the moment, the weather report for Saturday is excellent: Partly cloudy, highs in low 70's! Artur is planning a bike trail expedition. Bike rentals are available close by, so no need to bring yours if you don't want to. http://www.dcnr.state.pa.us/stateParks/parks/moraine.aspx People should start to arrive at 10 AM, and we'll have lunch and a presentation by Artur at NOON. Directions and car pooling info will follow this email. See you Saturday! Mike Why change of venue? Apparently due to the upcoming G-20, North Park is under 'administrative restriction' and they will not rent any pavilions until after G-20 is over. From mjbaysek at cs.cmu.edu Tue Sep 15 16:04:48 2009 From: mjbaysek at cs.cmu.edu (Michael J. Baysek) Date: Tue, 15 Sep 2009 16:04:48 -0400 Subject: [Research] LAB PICNIC RSVP/HEAD COUNT Message-ID: <4AAFF360.1080901@cs.cmu.edu> If you are planning to attend the picnic be sure to respond to each question below. This is to help with picnic planning. QUESTION SECTION ------------------------------------------------------ How many adults in your party: How many children: Dietary restrictions: Do you Need a ride from CMU? Can you offer a ride to anyone? END QUESTIONS ------------------------------------------------------- Again the time is 10 AM to 2 PM on this Saturday, Sep 19 at MORAINE STATE PARK. Bike (or kayak, etc) rentals will be reimbursed. If you want to do another activity, check the prices and let us know. [If you are not coming, sorry for the spam.] From mjbaysek at cs.cmu.edu Thu Sep 17 07:40:24 2009 From: mjbaysek at cs.cmu.edu (Michael J. Baysek) Date: Thu, 17 Sep 2009 07:40:24 -0400 Subject: [Research] Lab Picnic Directions Message-ID: <4AB22028.3010909@cs.cmu.edu> Here are directions from CMU to the area of the park where our pavilion is. http://maps.google.com/maps?f=d&source=s_d&saddr=5000+Forbes+Ave,+Pittsburgh,+PA+15213&daddr=I-79+N+to:40.962789,-80.111618&hl=en&geocode=FZIiaQIdECs8-w%3BFboJcQIdQj45-w%3B&mra=dme&mrcr=0&mrsp=2&sz=12&via=1&sll=40.965641,-80.130501&sspn=0.17136,0.329933&ie=UTF8&ll=40.71916,-80.04776&spn=0.687988,1.319733&z=10 This should put you in very close vicinity to the Lakeview #5 Pavilion. Picnic start time: 10AM Lunch at: NOON See you there! Mike -------------- next part -------------- An HTML attachment was scrubbed... URL: From awd at cs.cmu.edu Fri Sep 18 09:59:41 2009 From: awd at cs.cmu.edu (Artur Dubrawski) Date: Fri, 18 Sep 2009 09:59:41 -0400 Subject: [Research] new series of th Auton Lab meetings Message-ID: <4AB3924D.7080306@cs.cmu.edu> Dear Autonians, Would you please provide your availability information to Karen Widmaier (krw at andrew.cmu.edu, also cc-d with this) so that we could finalize scheduling of the Auton Lab meetings for this semester. The current assumption is that we can keep the meetings in the time slot between 11:30am and 1:30pm on Tuesdays with the fall-back on Thursdays. We may be able to change that if we see many conflicts. We plan the first meeting of the new series for the last week of September (probably Tue Sep 29th). Thanks very much, Artur From mjbaysek at cs.cmu.edu Fri Sep 18 17:52:59 2009 From: mjbaysek at cs.cmu.edu (Michael J. Baysek) Date: Fri, 18 Sep 2009 17:52:59 -0400 Subject: [Research] Picnic Stuff - See You There! Message-ID: <4AB4013B.3080500@cs.cmu.edu> Hi everyone. We hope to see you all at the picnic. We are planning to do excursions before the lunch. So if you are planning to go on bike trails or kayaking, be sure to be at the Lakeview #5 pavilion by 10 AM. We will be breaking into the groups and leaving from the Pavilion area shortly after 10. Lunch will be around NOON. My cell phone number is 412-596-1618. Please call me if you find reason to. See you there :) -Mike Directions. again: http://maps.google.com/maps?f=d&source=s_d&saddr=5000+Forbes+Ave,+Pittsburgh,+PA+15213&daddr=I-79+N+to:40.962789,-80.111618&hl=en&geocode=FZIiaQIdECs8-w%3BFboJcQIdQj45-w%3B&mra=dme&mrcr=0&mrsp=2&sz=12&via=1&sll=40.965641,-80.130501&sspn=0.17136,0.329933&ie=UTF8&ll=40.71916,-80.04776&spn=0.687988,1.319733&z=10 From awd at cs.cmu.edu Sun Sep 20 10:16:52 2009 From: awd at cs.cmu.edu (Artur Dubrawski) Date: Sun, 20 Sep 2009 10:16:52 -0400 Subject: [Research] It's been a great picnic! Message-ID: <4AB63954.8080006@cs.cmu.edu> Dear Autonians, I enjoyed the picnic very much. We've got to have them more often. Huge thanks to Mike and Robin for taking care of all the important details and for the great food and thanks to everyone who attended for making it such a pleasant and relaxing event. BTW, plans for the next Lab outdoors outing - an one-day hiking expedition in the near future (2-3 weekends from now) - are underway. Please let Purna and I know if you'd like to participate in planing. Artur From mailer at doodle.com Wed Sep 23 12:48:03 2009 From: mailer at doodle.com (Karen Widmaier via Doodle) Date: Wed, 23 Sep 2009 18:48:03 +0200 (CEST) Subject: [Research] AutonLab weekly meetings Message-ID: <14926825.413449.1253724483738.JavaMail.doodle@worker1> Karen Widmaier invites you to participate in Doodle poll "AutonLab weekly meetings". Follow this link to participate in it: http://www.doodle.com/rgvkwswp2afwnsyc ---- Advertisement -------------------------------- Are you a regular Doodle user? Use Doodle ads-free and with your personal design Order now: http://www.doodle.com/r/PremiumDoodleEN -------------------------------- Advertisement ---- From ajit at cs.cmu.edu Sun Sep 27 23:09:44 2009 From: ajit at cs.cmu.edu (Ajit Paul Singh) Date: Sun, 27 Sep 2009 22:09:44 -0500 Subject: [Research] Fwd: Thesis Defense - Ajit Singh - 9/30/09 References: <4AB28567.9080403@cs.cmu.edu> Message-ID: I'll be defending this Wednesday at noon. Everyone's welcome to attend (NSH 1507). -- Ajit Begin forwarded message: > From: Diane Stidle > Date: September 17, 2009 1:52:23 PM CDT > To: ml-seminar at cs.cmu.edu > Subject: Thesis Defense - Ajit Singh - 9/30/09 > > Thesis Defense > > Date: 9/30/09 > Time: 12:00pm > Place: 1507 NSH > > PhD Candidate: Ajit Singh > > Title: *Efficient Models for Relational Learning > * > Abstract: > Information integration deals with the setting where one has > multiple sources of data, each describing different properties of > the same set of entities. We are concerned primarily with settings > where the properties are pairwise relations between entities, and > attributes of entities. We want to predict the value of relations > and attributes, but relations between entities violate the basic > statistical assumption of exchangeable data points, or entities. > Furthermore, we desire models that scale gracefully as the number > of entities and relations increase. > > Matrices are the simplest form of relational data; and we begin by > distilling the literature on low-rank matrix factorization into a > small number of modelling choices. We then frame information > integration as simultaneously factoring sets of related matrices: > i.e., Collective Matrix Factorization. Each entity is described by > a small number of parameters, and if an entity is described by more > than one matrix, those parameters participate in multiple matrix > factorizations. Maximum likelihood estimation of the resulting > model involves a large non-convex optimization, which we reduce to > cyclically solving convex optimizations over small subsets of the > parameters. Each convex subproblem can be solved by Newton-Raphson, > which we extend to the setting of stochastic Newton-Raphson. > > To address the limitations of maximum likelihood estimation in > matrix factorization models, we extend our approach to the > hierarchical Bayesian setting. Here, Bayesian estimation involves > computing a high-dimensional integral with no analytic form. If we > resorted to standard Metropolis-Hastings techniques, slow mixing > would limit the scalability of our approach to large sets of > entities. We show how to accelerate Metropolis-Hastings by using > our efficient solution for maximum likelihood estimation to guide > the sampling process. > > This thesis rests on two claims, that (i) that Collective Matrix > Factorization can effectively integrate different sources of data > to improve prediction; and, (ii) that training scales well as the > number of entities and observations increase. Two real-world data > sets are considered in experimental support of these claims: > augmented collaborative filtering and augmented brain imaging. In > augmented collaborative filtering, we show that genre information > about movies can be used to increase the predictive accuracy of > user's ratings. In augmented brain imaging, we show that word co- > occurrence information can be used to increase the predictive > accuracy of a model of changes in brain activity to word stimuli, > even in regions of the brain that were never included in the > training data. > > http://www.cs.cmu.edu/~ajit/pubs/ajit-thesis-submitted.pdf > > Thesis Committee: > Geoff Gordon (chair) > Tom Mitchell > Christos Faloutsos > Pedro Domingos (University of Washington) > > -- > ******************************************************************* > Diane Stidle > Business & Graduate Programs Manager > Machine Learning Department > School of Computer Science > 8203 Gates Hillman Complex > Carnegie Mellon University > 5000 Forbes Avenue > Pittsburgh, PA 15213-3891 > Phone: 412-268-1299 > Fax: 412-268-3431 > Email: diane at cs.cmu.edu > URL:http://www.ml.cmu.edu > > -------------- next part -------------- An HTML attachment was scrubbed... URL: From awd at cs.cmu.edu Mon Sep 28 10:22:56 2009 From: awd at cs.cmu.edu (Artur Dubrawski) Date: Mon, 28 Sep 2009 10:22:56 -0400 Subject: [Research] Auton Lab meetings to resume on Wednesday October 7th Message-ID: <4AC0C6C0.3020802@cs.cmu.edu> Dear Autonians, It is time to get back to business. Wednesday noon has won, by a land slide, the popular vote for the lab meeting time for 2009/10. Thank you very much for participating in the survey. We will start the series on Wednesday next week. The topic and the exact location will be announced later. This Wednesday, please consider attending a thesis defense by a former Autonian, on the topic closely related to our ongoing area of interest (see below for details). Thanks Artur ----- Date: 9/30/09 Time: 12:00pm Place: 1507 NSH PhD Candidate: Ajit Singh Title: *Efficient Models for Relational Learning * Abstract: Information integration deals with the setting where one has multiple sources of data, each describing different properties of the same set of entities. We are concerned primarily with settings where the properties are pairwise relations between entities, and attributes of entities. We want to predict the value of relations and attributes, but relations between entities violate the basic statistical assumption of exchangeable data points, or entities. Furthermore, we desire models that scale gracefully as the number of entities and relations increase. Matrices are the simplest form of relational data; and we begin by distilling the literature on low-rank matrix factorization into a small number of modelling choices. We then frame information integration as simultaneously factoring sets of related matrices: i.e., Collective Matrix Factorization. Each entity is described by a small number of parameters, and if an entity is described by more than one matrix, those parameters participate in multiple matrix factorizations. Maximum likelihood estimation of the resulting model involves a large non-convex optimization, which we reduce to cyclically solving convex optimizations over small subsets of the parameters. Each convex subproblem can be solved by Newton-Raphson, which we extend to the setting of stochastic Newton-Raphson. To address the limitations of maximum likelihood estimation in matrix factorization models, we extend our approach to the hierarchical Bayesian setting. Here, Bayesian estimation involves computing a high-dimensional integral with no analytic form. If we resorted to standard Metropolis-Hastings techniques, slow mixing would limit the scalability of our approach to large sets of entities. We show how to accelerate Metropolis-Hastings by using our efficient solution for maximum likelihood estimation to guide the sampling process. This thesis rests on two claims, that (i) that Collective Matrix Factorization can effectively integrate different sources of data to improve prediction; and, (ii) that training scales well as the number of entities and observations increase. Two real-world data sets are considered in experimental support of these claims: augmented collaborative filtering and augmented brain imaging. In augmented collaborative filtering, we show that genre information about movies can be used to increase the predictive accuracy of user's ratings. In augmented brain imaging, we show that word co-occurrence information can be used to increase the predictive accuracy of a model of changes in brain activity to word stimuli, even in regions of the brain that were never included in the training data. http://www.cs.cmu.edu/~ajit/pubs/ajit-thesis-submitted.pdf Thesis Committee: Geoff Gordon (chair) Tom Mitchell Christos Faloutsos Pedro Domingos (University of Washington) From mjbaysek at cs.cmu.edu Wed Sep 30 15:52:31 2009 From: mjbaysek at cs.cmu.edu (Michael J. Baysek) Date: Wed, 30 Sep 2009 15:52:31 -0400 Subject: [Research] Lab Picnic Pictures Message-ID: <4AC3B6FF.5000403@cs.cmu.edu> Hi everyone. Saswati has taken pics from the lab picnic. I have posted these at the following location /mnt/BigPapa/home/public/lab_picnic_2009-09-12/ Enjoy!