From jhgood at cs.cmu.edu Tue Oct 1 10:59:38 2024 From: jhgood at cs.cmu.edu (Jack Good) Date: Tue, 1 Oct 2024 10:59:38 -0400 Subject: Full scratch directories Message-ID: Hi all, The scratch directories are full on lov8 and lov11 and close to full on lov7, lov9, lov10, and lov13. Please clear out any large files you no longer need. Thanks, Jack Good -------------- next part -------------- An HTML attachment was scrubbed... URL: From awd at cs.cmu.edu Tue Oct 1 15:54:12 2024 From: awd at cs.cmu.edu (Artur Dubrawski) Date: Tue, 1 Oct 2024 15:54:12 -0400 Subject: Full scratch directories In-Reply-To: References: Message-ID: Team, This is a part of basic etiquette of any user of the autonlab computing resources. Not following the rules almost always causes widespread harm. Thanks for your understanding and compliance, Artur On Tue, Oct 1, 2024 at 11:00?AM Jack Good wrote: > Hi all, > > The scratch directories are full on lov8 and lov11 and close to full on > lov7, lov9, lov10, and lov13. Please clear out any large files you no > longer need. > > Thanks, > Jack Good > -------------- next part -------------- An HTML attachment was scrubbed... URL: From jieshic at andrew.cmu.edu Wed Oct 2 10:53:30 2024 From: jieshic at andrew.cmu.edu (Jieshi Chen) Date: Wed, 2 Oct 2024 10:53:30 -0400 Subject: You're invited! Auton Lab 31st Anniversary Picnic - Sunday, October 13th @ Vietnam Veterans Pavilion In-Reply-To: References: Message-ID: Hello Everyone, Just a reminder of the picnic RSVP if you have not done so. Thank you! *https://forms.gle/864oaDofmTWpKPEC9 * Best, Jessie On Thu, Sep 19, 2024 at 1:48?PM Jieshi Chen wrote: > Dear Autonians, > > We are excited to invite you to celebrate the 31st anniversary of our lab! > To mark the occasion, we are hosting a family-friendly picnic, and we would > love for you to join us. This will be a great opportunity to relax, enjoy > good food, and celebrate with colleagues, friends, and family. Feel free to > bring along your loved ones ? everyone is welcome! > > *Date:* Sunday, October 13th > *Time:* Noon - 5pm (venue reserved from 11am to 9pm) > *Location:* Vietnam Veterans Pavilion at Schenley Park (Google Map > ) > > Please *RSVP *through the link below so we can make sure there's plenty > of food and fun for everyone. > > *https://forms.gle/864oaDofmTWpKPEC9 * > > Looking forward to celebrating with you! > > > Best, > Jessie > > -- > > Jieshi Chen > Principal Research Analyst > Auton Lab, Robotics Institute, School of Computer Science > Carnegie Mellon University > Email: jieshic at andrew.cmu.edu > Office: Newell Simon Hall Room 3115 > Address: 5000 Forbes Avenue, Pittsburgh, PA 15213 > -------------- next part -------------- An HTML attachment was scrubbed... URL: From jieshic at andrew.cmu.edu Tue Oct 8 10:36:25 2024 From: jieshic at andrew.cmu.edu (Jieshi Chen) Date: Tue, 8 Oct 2024 10:36:25 -0400 Subject: You're invited! Auton Lab 31st Anniversary Picnic - Sunday, October 13th @ Vietnam Veterans Pavilion In-Reply-To: References: Message-ID: Hello Everyone, The picnic is happening this coming Sunday!! Pls remember to RSVP if you have not done so. Thank you! *https://forms.gle/864oaDofmTWpKPEC9 * Best, Jessie On Thu, Sep 19, 2024 at 1:48?PM Jieshi Chen wrote: > Dear Autonians, > > We are excited to invite you to celebrate the 31st anniversary of our lab! > To mark the occasion, we are hosting a family-friendly picnic, and we would > love for you to join us. This will be a great opportunity to relax, enjoy > good food, and celebrate with colleagues, friends, and family. Feel free to > bring along your loved ones ? everyone is welcome! > > *Date:* Sunday, October 13th > *Time:* Noon - 5pm (venue reserved from 11am to 9pm) > *Location:* Vietnam Veterans Pavilion at Schenley Park (Google Map > ) > > Please *RSVP *through the link below so we can make sure there's plenty > of food and fun for everyone. > > *https://forms.gle/864oaDofmTWpKPEC9 * > > Looking forward to celebrating with you! > > > Best, > Jessie > > -- > > Jieshi Chen > Principal Research Analyst > Auton Lab, Robotics Institute, School of Computer Science > Carnegie Mellon University > Email: jieshic at andrew.cmu.edu > Office: Newell Simon Hall Room 3115 > Address: 5000 Forbes Avenue, Pittsburgh, PA 15213 > -------------- next part -------------- An HTML attachment was scrubbed... URL: From awd at cs.cmu.edu Thu Oct 10 16:25:24 2024 From: awd at cs.cmu.edu (Artur Dubrawski) Date: Thu, 10 Oct 2024 16:25:24 -0400 Subject: unprecedented acceptance rate for Auton workshop papers Message-ID: Team, I am very pleased to share that we just had 7 successes in 7 attempts with our NeurIPS (6) and ICAIF (1) workshop papers submitted on topics surrounding foundation models. We sometimes get a 100% hit rate but it does not happen 100% of the time. I am particularly impressed with the papers led by our recent summer interns included in the above count. Congrats Willa, Nina, Michal, Yifu, Arjun, Christian. and Mononito! Cheers Artur -------------- next part -------------- An HTML attachment was scrubbed... URL: From kelenber at andrew.cmu.edu Thu Oct 10 16:28:37 2024 From: kelenber at andrew.cmu.edu (Kimberly Elenberg) Date: Thu, 10 Oct 2024 16:28:37 -0400 Subject: unprecedented acceptance rate for Auton workshop papers In-Reply-To: References: Message-ID: AMAZING and honestly- a great reflection on the students and the team! Best regards, Kimberly On Thu, Oct 10, 2024 at 4:26?PM Artur Dubrawski wrote: > Team, > > I am very pleased to share that we just had 7 successes in 7 attempts with > our NeurIPS (6) and ICAIF (1) workshop papers submitted on topics > surrounding foundation models. > > We sometimes get a 100% hit rate but it does not happen 100% of the time. > > I am particularly impressed with the papers led by our recent summer > interns included in the above count. > > Congrats Willa, Nina, Michal, Yifu, Arjun, Christian. and Mononito! > > Cheers > Artur > -------------- next part -------------- An HTML attachment was scrubbed... URL: From rburcin at andrew.cmu.edu Thu Oct 10 16:32:18 2024 From: rburcin at andrew.cmu.edu (Rachel Burcin) Date: Thu, 10 Oct 2024 16:32:18 -0400 Subject: unprecedented acceptance rate for Auton workshop papers In-Reply-To: References: Message-ID: Amazing! Congratulations to each of you. The Auton Lab is definitely a special place. ? Rachel Burcin rachel at cmu.edu On Thu, Oct 10, 2024 at 4:28?PM Kimberly Elenberg wrote: > AMAZING and honestly- a great reflection on the students and the team! > > Best regards, > Kimberly > > On Thu, Oct 10, 2024 at 4:26?PM Artur Dubrawski wrote: > >> Team, >> >> I am very pleased to share that we just had 7 successes in 7 attempts >> with our NeurIPS (6) and ICAIF (1) workshop papers submitted on topics >> surrounding foundation models. >> >> We sometimes get a 100% hit rate but it does not happen 100% of the time. >> >> I am particularly impressed with the papers led by our recent summer >> interns included in the above count. >> >> Congrats Willa, Nina, Michal, Yifu, Arjun, Christian. and Mononito! >> >> Cheers >> Artur >> > -------------- next part -------------- An HTML attachment was scrubbed... URL: From awd at cs.cmu.edu Thu Oct 10 16:46:01 2024 From: awd at cs.cmu.edu (Artur Dubrawski) Date: Thu, 10 Oct 2024 16:46:01 -0400 Subject: CORRECTION: unprecedented acceptance rate for Auton workshop papers In-Reply-To: References: Message-ID: I was too quick to type! A few corrections: 1.We had 8 (eight) not 7 (seven) NeurIPS workshop papers submitted by the below group of authors, and accepted. The NeurIPS workshop paper acceptance rate success rate remains NeurIPS workshop acceptance rate remains at 100%. 2. I forgot to add Rachel's name to the list of the Auton authors of those successful papers. I am sorry! 3. My spell corrector has butchered Cristian's name and I overlooked that. I am sorry! 4. The ICAIF paper is not yet technically accepted, stand by for more good news, hopefully. Artur On Thu, Oct 10, 2024 at 4:25?PM Artur Dubrawski wrote: > Team, > > I am very pleased to share that we just had 7 successes in 7 attempts with > our NeurIPS (6) and ICAIF (1) workshop papers submitted on topics > surrounding foundation models. > > We sometimes get a 100% hit rate but it does not happen 100% of the time. > > I am particularly impressed with the papers led by our recent summer > interns included in the above count. > > Congrats Willa, Nina, Michal, Yifu, Arjun, Christian. and Mononito! > > Cheers > Artur > -------------- next part -------------- An HTML attachment was scrubbed... URL: From kdgutier at andrew.cmu.edu Thu Oct 10 19:11:00 2024 From: kdgutier at andrew.cmu.edu (Kin Gutierrez Olivares) Date: Thu, 10 Oct 2024 19:11:00 -0400 Subject: unprecedented acceptance rate for Auton workshop papers In-Reply-To: References: Message-ID: Add to 7/7 another 2/2 from my side. Go AutonLab ? On Thu, Oct 10, 2024 at 4:33?PM Rachel Burcin wrote: > Amazing! Congratulations to each of you. The Auton Lab is definitely a > special place. ? > > Rachel Burcin > rachel at cmu.edu > > > On Thu, Oct 10, 2024 at 4:28?PM Kimberly Elenberg > wrote: > >> AMAZING and honestly- a great reflection on the students and the team! >> >> Best regards, >> Kimberly >> >> On Thu, Oct 10, 2024 at 4:26?PM Artur Dubrawski wrote: >> >>> Team, >>> >>> I am very pleased to share that we just had 7 successes in 7 attempts >>> with our NeurIPS (6) and ICAIF (1) workshop papers submitted on topics >>> surrounding foundation models. >>> >>> We sometimes get a 100% hit rate but it does not happen 100% of the time. >>> >>> I am particularly impressed with the papers led by our recent summer >>> interns included in the above count. >>> >>> Congrats Willa, Nina, Michal, Yifu, Arjun, Christian. and Mononito! >>> >>> Cheers >>> Artur >>> >> -------------- next part -------------- An HTML attachment was scrubbed... URL: From awd at cs.cmu.edu Thu Oct 10 20:17:11 2024 From: awd at cs.cmu.edu (Artur Dubrawski) Date: Thu, 10 Oct 2024 20:17:11 -0400 Subject: unprecedented acceptance rate for Auton workshop papers In-Reply-To: References: Message-ID: Ok, easy now, have there been any non-Auton papers accepted to NeurIPS workshops in the current batch? :P Congrats Kin! Artur On Thu, Oct 10, 2024 at 7:11?PM Kin Gutierrez Olivares < kdgutier at andrew.cmu.edu> wrote: > Add to 7/7 another 2/2 from my side. > > Go AutonLab ? > > > On Thu, Oct 10, 2024 at 4:33?PM Rachel Burcin > wrote: > >> Amazing! Congratulations to each of you. The Auton Lab is definitely a >> special place. ? >> >> Rachel Burcin >> rachel at cmu.edu >> >> >> On Thu, Oct 10, 2024 at 4:28?PM Kimberly Elenberg < >> kelenber at andrew.cmu.edu> wrote: >> >>> AMAZING and honestly- a great reflection on the students and the team! >>> >>> Best regards, >>> Kimberly >>> >>> On Thu, Oct 10, 2024 at 4:26?PM Artur Dubrawski wrote: >>> >>>> Team, >>>> >>>> I am very pleased to share that we just had 7 successes in 7 attempts >>>> with our NeurIPS (6) and ICAIF (1) workshop papers submitted on topics >>>> surrounding foundation models. >>>> >>>> We sometimes get a 100% hit rate but it does not happen 100% of the >>>> time. >>>> >>>> I am particularly impressed with the papers led by our recent summer >>>> interns included in the above count. >>>> >>>> Congrats Willa, Nina, Michal, Yifu, Arjun, Christian. and Mononito! >>>> >>>> Cheers >>>> Artur >>>> >>> -------------- next part -------------- An HTML attachment was scrubbed... URL: From jieshic at andrew.cmu.edu Sat Oct 12 17:11:21 2024 From: jieshic at andrew.cmu.edu (Jieshi Chen) Date: Sat, 12 Oct 2024 17:11:21 -0400 Subject: You're invited! Auton Lab 31st Anniversary Picnic - Sunday, October 13th @ Vietnam Veterans Pavilion In-Reply-To: References: Message-ID: Hi Autonians, Just a reminder of tomorrow's picnic at 12pm at the Vietnam Veterans Pavilion at Schenley Park! Looking forward to seeing you! Best, Jessie On Thu, Sep 19, 2024 at 1:48?PM Jieshi Chen wrote: > Dear Autonians, > > We are excited to invite you to celebrate the 31st anniversary of our lab! > To mark the occasion, we are hosting a family-friendly picnic, and we would > love for you to join us. This will be a great opportunity to relax, enjoy > good food, and celebrate with colleagues, friends, and family. Feel free to > bring along your loved ones ? everyone is welcome! > > *Date:* Sunday, October 13th > *Time:* Noon - 5pm (venue reserved from 11am to 9pm) > *Location:* Vietnam Veterans Pavilion at Schenley Park (Google Map > ) > > Please *RSVP *through the link below so we can make sure there's plenty > of food and fun for everyone. > > *https://forms.gle/864oaDofmTWpKPEC9 * > > Looking forward to celebrating with you! > > > Best, > Jessie > > -- > > Jieshi Chen > Principal Research Analyst > Auton Lab, Robotics Institute, School of Computer Science > Carnegie Mellon University > Email: jieshic at andrew.cmu.edu > Office: Newell Simon Hall Room 3115 > Address: 5000 Forbes Avenue, Pittsburgh, PA 15213 > -------------- next part -------------- An HTML attachment was scrubbed... URL: From awd at cs.cmu.edu Mon Oct 14 11:24:55 2024 From: awd at cs.cmu.edu (Artur Dubrawski) Date: Mon, 14 Oct 2024 11:24:55 -0400 Subject: Fwd: RI Ph.D. Thesis Defense: Jack Good In-Reply-To: References: Message-ID: Please mark your calendars to attend this transformative event! Cheers Artur ---------- Forwarded message --------- From: Suzanne Muth Date: Mon, Oct 14, 2024 at 11:06?AM Subject: RI Ph.D. Thesis Defense: Jack Good To: RI People *Date:* 23 October 2024 *Time:* 3:30 p.m. (ET) *Location:* GHC 6501 *Zoom Link:* https://cmu.zoom.us/j/98137707124?pwd=g1OvJBlbgfZFtLQm3tIaEWos9eFhcZ.1 *Type:* Ph.D. Thesis Defense *Who:* Jack Good *Title:* Trustworthy Learning using Uncertain Interpretation of Data *Abstract:* Motivated by the potential of Artificial Intelligence (AI) in high-cost and safety-critical applications, and recently also by the increasing presence of AI in our everyday lives, Trustworthy AI has grown in prominence as a broad area of research encompassing topics such as interpretability, robustness, verifiable safety, fairness, privacy, accountability, and more. This has created a tension between simple, transparent models with inherent trust-related benefits and complex, black-box models with unparalleled performance on many tasks. Towards closing this gap, we propose and study an uncertain interpretation of numerical data and apply it to tree-based models, resulting in a novel kind of fuzzy decision tree called Kernel Density Decision Trees (KDDTs) with improved performance, enhanced trustworthy qualities, and increased utility, enabling the use of these trees in broader applications. We group the contributions of this thesis into three pillars. The first pillar is robustness and verification. The uncertain interpretation, by accounting for uncertainty in the data, and more generally as a kind of regularization on the function represented by a model, can improve the model with respect to various notions of robustness. We demonstrate its ability to improve robustness to noisy features and noisy labels, both of which are common in real-world data. Next, we show how efficiently verifiable adversarial robustness is achievable through the theory of randomized smoothing. Finally, we discuss the related topic of verification and propose the first verification algorithm for fuzzy decision trees. The second pillar is interpretability. While decision trees are widely considered to be interpretable, good performance from tree-based models is often limited to tabular data and demands both feature engineering, which increases design effort, and ensemble methods, which severely diminish interpretability compared to single-tree models. By leveraging the efficient fitting and differentiability of KDDTs, we propose a system of learning parameterized feature transformations for decision trees. By choosing interpretable feature classes and applying sparsity regularization,we can obtain compact single-tree models with competitive performance. We demonstrate application to tabular, time series, and simple image data. The third pillar is pragmatic advancements. Semi-supervised Learning (SSL) is motivated by the expense of labeling and learns from a mix of labeled and unlabeled data. SSL for trees is generally limited to black-box wrapper methods, for which trees are not well-suited. We propose as an alternative a novel intrinsic SSL method based on our uncertain interpretation of data. Federated Learning (FL) is motivated by data sharing limitations and learns from distributed data by communicating models. We introduce a new FL algorithm based on function space regularization, which borrows concepts and methods from our formalism of uncertain interpretation. Unlike prior FL methods, it supports non-parametric models and has convergence guarantees under mild assumptions. Finally, we show how our FL algorithm also provides a simple utility for ensemble merging. *Thesis Committee Members:* Artur Dubrawski, Chair Jeff Schneider Tom Mitchell Gilles Clermont, University of Pittsburgh A draft of the thesis defense document is available here . -------------- next part -------------- An HTML attachment was scrubbed... URL: From awd at cs.cmu.edu Wed Oct 16 12:44:51 2024 From: awd at cs.cmu.edu (Artur Dubrawski) Date: Wed, 16 Oct 2024 12:44:51 -0400 Subject: Rachel Burcin wins Andy Award! Message-ID: Autonians, Typically, I brag here about achievements of either current or past, but formal Auton Lab members. Today may feel different, but it is not, really. Rachel Burcin is definitely not just one of the best friends of the Auton Lab but obviously a member of the Autonian family, whether she is formally part of our roster or not. BTW, I do not know anyone who would deserve the award for Commitment to Students more than Rachel. Any student who ever came in contact with her could easily confirm. This recognition is well deserved and massively overdue. Please join me in sending the warmest congratulations to the one and only Rachel Burcin! Cheers, Artur PS Here is the synopsis of the award from www.cmu.edu/andyawards/nominees/commitment-to-students.html : *The Andy Award for Commitment to Students honors staff members who are dedicated to guiding and assisting university students. They surpass the duties of their job description to meet students' needs, serving as role models who display a positive and professional image at all times.Criteria include investing extra effort in satisfying student needs and expectations, identifying opportunities that support student needs and involving students in campus life beyond the classroom to emphasize their value as members of the university community.* PSPS The description does not provide enough detail to give Rachel full justice. One thing that is obviously missing there, or maybe it should be an entirely separate award that she also deserves, is that Rachel's heart is so huge it does not quite fit its normal spot in her body, so it spills over to take 100% of the volumes where other people keep their egoes. -------------- next part -------------- An HTML attachment was scrubbed... URL: From kelenber at andrew.cmu.edu Wed Oct 16 16:57:25 2024 From: kelenber at andrew.cmu.edu (Kimberly Elenberg) Date: Wed, 16 Oct 2024 16:57:25 -0400 Subject: Rachel Burcin wins Andy Award! In-Reply-To: References: Message-ID: Yes! I couldn?t think of a more deserving person for this recognition!!! Congratulations Rachel!!!! On Wed, Oct 16, 2024 at 12:45?PM Artur Dubrawski wrote: > Autonians, > > Typically, I brag here about achievements of either current or past, but > formal Auton Lab members. > > Today may feel different, but it is not, really. Rachel Burcin is > definitely not just one of the best friends of the Auton Lab but obviously > a member of the Autonian family, whether she is formally part of our roster > or not. > > BTW, I do not know anyone who would deserve the award for Commitment to > Students more than Rachel. Any student who ever came in contact with her > could easily confirm. This recognition is well deserved and massively > overdue. > > Please join me in sending the warmest congratulations to the one and only > Rachel Burcin! > > Cheers, > Artur > > PS Here is the synopsis of the award from > www.cmu.edu/andyawards/nominees/commitment-to-students.html : > > > > *The Andy Award for Commitment to Students honors staff members who are > dedicated to guiding and assisting university students. They surpass the > duties of their job description to meet students' needs, serving as role > models who display a positive and professional image at all times.Criteria > include investing extra effort in satisfying student needs and > expectations, identifying opportunities that support student needs and > involving students in campus life beyond the classroom to emphasize their > value as members of the university community.* > > PSPS The description does not provide enough detail to give Rachel full > justice. One thing that is obviously missing there, or maybe it should be > an entirely separate award that she also deserves, is that Rachel's heart > is so huge it does not quite fit its normal spot in her body, so it spills > over to take 100% of the volumes where other people keep their egoes. > -------------- next part -------------- An HTML attachment was scrubbed... URL: From awd at cs.cmu.edu Fri Oct 18 15:12:14 2024 From: awd at cs.cmu.edu (Artur Dubrawski) Date: Fri, 18 Oct 2024 15:12:14 -0400 Subject: Fwd: [Special Talk] Christos Faloutsos Presenting "Lessons Learned from 1,000 Research Paper Rejections" In-Reply-To: References:

Message-ID: I'd always recommend learning from someone else's mistakes (though I wish I followed that myself more often than not :P ). This talk is offering exactly that! Cheers Artur ---------- Forwarded message --------- From: Diane Stidle Date: Fri, Oct 18, 2024 at 2:59?PM Subject: Fwd: [Special Talk] Christos Faloutsos Presenting "Lessons Learned from 1,000 Research Paper Rejections" To: This looks like a really interesting talk! *Special Talk* *CHRISTOS FALOUTSOS* Thursday, November 14, 2024 3:30 PM - GHC 6115 *Title:* *Lessons Learned from 1,000 Research Paper Rejections* *Description:* What do we need to do to polish the write-up of our research papers? Papers that are otherwise worth publishing often get rejected due to presentation pitfalls. After about 1,000 rejections over the past 30+ years, some patterns started to emerge. In this talk, we discuss the three most important of these patterns, their potential remedies, as well as a list of an additional dozen of patterns and remedies. *Please register to attend*: https://forms.gle/6rsJqT7epz6V8pD29. *Full details can be found here*: https://csd.cmu.edu/calendar/special-talk-christos-faloutsos. We hope to see you there! *Matthew Stewart* Doctoral Programs Manager GHC 6021 | matthewstewart at cmu.edu -- -- Diane Stidle PhD Program Manager Machine Learning Department Carnegie Mellon University diane at cs.cmu.edu -------------- next part -------------- An HTML attachment was scrubbed... URL: From awd at cs.cmu.edu Tue Oct 22 10:24:13 2024 From: awd at cs.cmu.edu (Artur Dubrawski) Date: Tue, 22 Oct 2024 10:24:13 -0400 Subject: IT IS TOMORROW: RI Ph.D. Thesis Defense: Jack Good In-Reply-To: References: Message-ID: Please make sure to witness the transformation of Jack Good to Doctor Good tomorrow! Artur On Mon, Oct 14, 2024 at 11:24?AM Artur Dubrawski wrote: > Please mark your calendars to attend this transformative event! > > Cheers > Artur > > ---------- Forwarded message --------- > From: Suzanne Muth > Date: Mon, Oct 14, 2024 at 11:06?AM > Subject: RI Ph.D. Thesis Defense: Jack Good > To: RI People > > > *Date:* 23 October 2024 > *Time:* 3:30 p.m. (ET) > *Location:* GHC 6501 > *Zoom Link:* > https://cmu.zoom.us/j/98137707124?pwd=g1OvJBlbgfZFtLQm3tIaEWos9eFhcZ.1 > *Type:* Ph.D. Thesis Defense > *Who:* Jack Good > *Title:* Trustworthy Learning using Uncertain Interpretation of Data > > *Abstract:* > Motivated by the potential of Artificial Intelligence (AI) in high-cost > and safety-critical > applications, and recently also by the increasing presence of AI in our > everyday lives, Trustworthy AI has grown in prominence as a broad area of > research encompassing topics such as interpretability, robustness, > verifiable safety, fairness, privacy, accountability, and more. This has > created a tension between simple, transparent models with inherent > trust-related benefits and complex, black-box models with unparalleled > performance on many tasks. Towards closing this gap, we propose and study > an uncertain interpretation of numerical data and apply it to tree-based > models, resulting in a novel kind of fuzzy decision tree called Kernel > Density Decision Trees (KDDTs) with improved performance, enhanced > trustworthy qualities, and increased utility, enabling the use of these > trees in broader applications. We group the contributions of this thesis > into three pillars. > > The first pillar is robustness and verification. The uncertain > interpretation, by accounting for uncertainty in the data, and more > generally as a kind of regularization on the function represented by a > model, can improve the model with respect to various notions of robustness. > We demonstrate its ability to improve robustness to noisy features and > noisy labels, both of which are common in real-world data. Next, we show > how efficiently verifiable adversarial robustness is achievable through the > theory of randomized smoothing. Finally, we discuss the related topic of > verification and propose the first verification algorithm for fuzzy > decision trees. > > The second pillar is interpretability. While decision trees are widely > considered to be interpretable, good performance from tree-based models is > often limited to tabular data and demands both feature engineering, which > increases design effort, and ensemble methods, which severely diminish > interpretability compared to single-tree models. By leveraging the > efficient fitting and differentiability of KDDTs, we propose a system of > learning parameterized feature transformations for decision trees. By > choosing interpretable feature classes and applying sparsity > regularization,we can obtain compact single-tree models with competitive > performance. We demonstrate application to tabular, time series, and simple > image data. > > The third pillar is pragmatic advancements. Semi-supervised Learning (SSL) > is motivated by the expense of labeling and learns from a mix of labeled > and unlabeled data. SSL for trees is generally limited to black-box wrapper > methods, for which trees are not well-suited. We propose as an alternative > a novel intrinsic SSL method based on our uncertain interpretation of data. > Federated Learning (FL) is motivated by data sharing limitations and learns > from distributed data by communicating models. We introduce a new FL > algorithm based on function space regularization, which borrows concepts > and methods from our formalism of uncertain interpretation. Unlike prior FL > methods, it supports non-parametric models and has convergence guarantees > under mild assumptions. Finally, we show how our FL algorithm also provides > a simple utility for ensemble merging. > > *Thesis Committee Members:* > Artur Dubrawski, Chair > Jeff Schneider > Tom Mitchell > Gilles Clermont, University of Pittsburgh > > A draft of the thesis defense document is available here > > . > -------------- next part -------------- An HTML attachment was scrubbed... URL: From ndeka at cs.cmu.edu Thu Oct 24 13:59:10 2024 From: ndeka at cs.cmu.edu (Namrata Deka) Date: Thu, 24 Oct 2024 13:59:10 -0400 Subject: No space left on scratch and extra_scratch on gpu26 Message-ID: Hello Auton users, If you have old data/checkpoints/big files on extra_scratch and scratch in gpu26 could you please delete them ASAP? Thanks! *Best,* *Namrata* -------------- next part -------------- An HTML attachment was scrubbed... URL: From awd at cs.cmu.edu Mon Oct 28 11:07:19 2024 From: awd at cs.cmu.edu (Artur Dubrawski) Date: Mon, 28 Oct 2024 11:07:19 -0400 Subject: Vincent Jeansalme is a Doctor now :) In-Reply-To: <3CA75D99-A053-4E12-9DD2-99B8A0236A96@gmail.com> References: <3CA75D99-A053-4E12-9DD2-99B8A0236A96@gmail.com> Message-ID: Team, A former Autonian known to many of us as Vincent should from now on be addressed only as Dr. Vincent please. I am pleased. Congrats Dr. Vincent! Artur -------------- next part -------------- An HTML attachment was scrubbed... URL: -------------- next part -------------- A non-text attachment was scrubbed... Name: image1.jpeg Type: image/jpeg Size: 336947 bytes Desc: not available URL: From predragp at andrew.cmu.edu Mon Oct 28 12:36:05 2024 From: predragp at andrew.cmu.edu (Predrag Punosevac) Date: Mon, 28 Oct 2024 10:36:05 -0600 Subject: Vincent Jeansalme is a Doctor now :) In-Reply-To: References: <3CA75D99-A053-4E12-9DD2-99B8A0236A96@gmail.com> Message-ID: And not to be forgotten the lady standing next to him is also Autonian! On Mon, Oct 28, 2024, 9:11 AM Artur Dubrawski wrote: > Team, > > A former Autonian known to many of us as Vincent should from now on be > addressed only as Dr. Vincent please. > > I am pleased. > > Congrats Dr. Vincent! > > Artur > > -------------- next part -------------- An HTML attachment was scrubbed... URL: