@LifelongML_Penn Profile picture

LifelongML

@LifelongML_Penn

We develop methods for AI systems to learn over long-term deployments, with applications primarily to robotics and medicine. PI: Eric Eaton, PhD @GRASPlab

Similar User
ContinualAI photo

@ContinualAI

CoLLAs 2025 photo

@CoLLAs_Conf

Kyle Vedder photo

@KyleVedder

Rahaf Aljundi photo

@AljundiRahaf

Dinesh Jayaraman photo

@dineshjayaraman

Jason Ma photo

@JasonMa2020

Tyler Hayes photo

@TylerLHayes

Jaekyeom Kim @ EMNLP photo

@Jaekyeom__Kim

Xingchen Wan photo

@wanxingchen_

Grégoire Delétang photo

@gregdeletang

Benjamin Thérien photo

@benjamintherien

Sirui Xu photo

@xu_sirui

Kaustab Pal photo

@kaustabpal

umberto michieli photo

@umbertomichieli

Aswin Raghavan photo

@aswin_raghavan_

LifelongML Reposted

📢 Exciting News! The Fourth Conference on Lifelong Learning Agents (CoLLAs 2025) will be held at the University of Pennsylvania (@Penn) in Philadelphia, USA 🇺🇸 🗓️ Important Dates: Abstract Deadline: Feb 21, 2025 Submission Deadline: Feb 26, 2025 Conference Dates: Aug 11 - Aug…

Tweet Image 1

LifelongML Reposted

This is the essence of Continual Learning. The answer seems to be yes. arxiv.org/abs/2205.12393


Exciting work from @MetaAI looking at LLMs from the perspective of continual learners, demonstrating learning without forgetting across a range of datasets! Is this style of model the future of continual learning? twitter.com/MetaAI/status/…

Continual-T0 (CT0) displays Continual Learning capabilities via self-supervision. This fine-tuned language model retains skills while learning new tasks across an unprecedented scale of 70 datasets. It can even combine instructions without prior training. bit.ly/3zhZJ32

Tweet Image 1


Great talk today by @Laparoscopes about the future of data science in surgery and how engineers / scientists can help build data driven systems to produce better patient outcomes!


Congratulations to @GummadiMeghna for her first paper with our group! Don't miss her poster presentation today @CoLLAs_Conf at 11 a.m. ET on task-agnostic #ContinualLearning via novelty detection.

Excited to share that our work at @LifelongML_Penn @GRASPlab on continual learning without class boundaries has been accepted to @CoLLAs_Conf In this work we introduce Sparse-High-level-Exclusive, Low-level-Shared feature representation (SHELS) for seamless #ContinualLearning.



LifelongML Reposted

How can a #MachineLearning agent reuse knowledge over a lifetime of tasks? Check out our new survey on #ContinualLearning and #Composition. We explore these two (~disjoint) fields and draw connections between them, offering avenues for future work. 🧵👇 arxiv.org/pdf/2207.07730…

Tweet Image 1

LifelongML Reposted

We are happy to announce the release of #CompoSuite, our benchmark of #robotic manipulation tasks for studying functional #compositionality for #MultiTaskLearning and #ContinualLearning in #ReinforcementLearning! @CoLLAs_Conf 🧵👇 arxiv.org/pdf/2207.04136… github.com/Lifelong-ML/Co…


Check out our newly released benchmark on #Compositional #ReinforcementLearning. Also, congratulations to @marcel_hussing on his first paper with us!

I'm delighted to share that my first paper at @LifelongML_Penn has been accepted to @CoLLAs_Conf 🥳. In it, we present CompoSuite: A Compositional Reinforcement Learning Benchmark as a testbed to evaluate the compositional capabilities of #ReinforcementLearning algorithms 🧵👇.



Our paper on Sparse PointPillars was accepted to #IROS2022! Check out @KyleVedder’s thread for details on the method. arXiv: arxiv.org/abs/2106.06882 Website: vedder.io/sparse_point_p…

I'm excited to announce Sparse PointPillars has been accepted to #iros2022 I'm looking forward to seeing everyone in Kyoto, Japan in October!



LifelongML Reposted

I gave an invited talk about Sparse PointPillars at the 3D-Deep Learning for Automated Driving workshop of IV2022. If you're interested in 3D object detection for real world systems, give it a watch! youtube.com/watch?v=JgcR6c…

Sparse PointPillars is a 3D object detector that runs efficiently on service robot compute platforms. How does it work? By maintaining and exploiting input sparsity in its pipeline, it performs many fewer operations compared to its dense counterpart, PointPillars.



If you are at @ieee_ras_icra today, come to the @GRASPlab technical tour 9:30-12:30 and check out our *real* robot continually learning to predict a changing room layout via occupancy map prediction! Part of an ongoing collaboration with @KostasPenn's group. #ICRA #ICRA2022


LifelongML Reposted

I will be presenting this work virtually in just a couple of hours (1:30pm ET) at #ICLR2022. Stop by to chat if you’re interested in #ContinualLearning #Compositionality #Modularity #ReinforcementLearning!

#iclr2022 Excited to share that our work on modular lifelong RL was accepted to @iclr_conf! 1/5 openreview.net/pdf?id=5XmLzds…



LifelongML Reposted

Huge congratulations to @JorgeAMendez_ , who successfully defended his dissertation "Lifelong Machine Learning of Functionally Compositional Structures" today at @Penn Absolutely groundbreaking work. Go Jorge!!! #continuallearning #MachineLearning @CIS_Penn


Check out our latest work on an efficient 3D object detector for embedded systems such as our service robots! twitter.com/KyleVedder/sta…

Tweet Image 1

Sparse PointPillars is a 3D object detector that runs efficiently on service robot compute platforms. How does it work? By maintaining and exploiting input sparsity in its pipeline, it performs many fewer operations compared to its dense counterpart, PointPillars.



Check out some of the latest work from our group by @JorgeAMendez_, which just got accepted into ICLR 2022. Keep an eye out for the camera-ready version soon!

#iclr2022 Excited to share that our work on modular lifelong RL was accepted to @iclr_conf! 1/5 openreview.net/pdf?id=5XmLzds…



Check out this exciting initiative!

We are thrilled to announce a new focused #MachineLearning conference: International Conference on Lifelong Learning Agents (CoLLA) lifelong-ml.cc 1/

Tweet Image 1


LifelongML Reposted

Today's talk in the #RobotLearning @NeurIPSConf countdown is by @GRASPlab's Eric Eaton (seas.upenn.edu/~eeaton/) & Jorge Mendez (@JorgeAMendez_) on "Factorized and Composable Representations for Lifelong Learning" Check it out! youtube.com/watch?v=JIGF0J…


Loading...

Something went wrong.


Something went wrong.