BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Penn Engineering Events - ECPv6.15.18//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-ORIGINAL-URL:https://seasevents.nmsdev7.com
X-WR-CALDESC:Events for Penn Engineering Events
REFRESH-INTERVAL;VALUE=DURATION:PT1H
X-Robots-Tag:noindex
X-PUBLISHED-TTL:PT1H
BEGIN:VTIMEZONE
TZID:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:20240310T070000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:20241103T060000
END:STANDARD
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:20250309T070000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:20251102T060000
END:STANDARD
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:20260308T070000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:20261101T060000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250404T090000
DTEND;TZID=America/New_York:20250404T160000
DTSTAMP:20260403T160431
CREATED:20250328T181917Z
LAST-MODIFIED:20250328T181917Z
UID:10008337-1743757200-1743782400@seasevents.nmsdev7.com
SUMMARY:Penn AI Governance Workshop
DESCRIPTION:This event is presented by the Wharton Accountable AI Lab and co-sponsored by the Wharton AI & Analytics Initiative; Penn Engineering; the Center for Technology\, Innovation & Competition; and the Perry World House. \nThe Penn AI Governance Workshop will feature panel discussions\, lightning talks and networking opportunities\, and will conclude with a reception. \nThe workshop brings together leading Penn researchers to explore key topics\, including AI safety and evaluation\, AI policy\, AI ethics and responsibility\, and teaching AI governance at Penn. Throughout the day\, panel discussions will be interspersed with lightning talks from researchers.
URL:https://seasevents.nmsdev7.com/event/penn-ai-governance-workshop/
LOCATION:Perry World House\, 3803 Locust Walk\, Philadelphia\, PA\, 19104\, United States
CATEGORIES:AI Month
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250404T093000
DTEND;TZID=America/New_York:20250404T103000
DTSTAMP:20260403T160431
CREATED:20250327T203720Z
LAST-MODIFIED:20250327T203720Z
UID:10008333-1743759000-1743762600@seasevents.nmsdev7.com
SUMMARY:CIS Seminar: "Pareto-efficient AI systems: Expanding the quality and efficiency frontier of AI"
DESCRIPTION:We have made exciting progress in AI by massive models on massive amounts of data center compute. However\, the demands for AI are rapidly expanding. I identify how to maximize performance under any compute constraint\, expanding the Pareto frontier of AI capabilities.\n \nThis talk builds up to an efficient language model architecture that expands the Pareto-frontier between quality and throughput efficiency. In motivation\, the Transformer\, AI’s current workhorse architecture\, is memory hungry\, severely limiting its throughput\, or amount of text it can process per second. This has led to a Cambrian explosion of alternate efficient architecture candidates proposed across prior work. Prior work has painted an exciting picture: there exists architectures that are asymptotically faster than Transformers\, while also matching quality. However\, I ask\, if we’re using asymptotically faster building blocks\, are we giving something up in quality?\n\n\nIn part one\, we build understanding. Indeed\, there’s no free lunch! I present my work to identify and explain the fundamental quality and efficiency tradeoffs between different classes of architectures. Methods I developed for this analysis are now ubiquitous in the development of language models.\n\n\nIn part two\, we measure how AI architecture candidates fare on the tradeoff space. A major hurdle\, however\, is that we lack implementations of the architectures that that run at peak-efficiency on modern hardware. Further\, many proposed architectures are asymptotically fast\, but not wall-clock fast. I present ThunderKittens\, a new programming library I built to help AI researchers write simple\, hardware-efficient algorithms across hardware platforms.\n\n\nIn part three\, we expand the Pareto-frontier of the tradeoff space. I present the BASED architecture\, which is built from simple\, hardware-efficient components. I released the state-of-the-art 8B-405B Transformer-free language models\, per standard evaluations\, all on an academic budget.\n\n\nGiven the massive investment into language models\, this work has had significant impact and adoption in research\, open-source\, and industry
URL:https://seasevents.nmsdev7.com/event/cis-seminar-pareto-efficient-ai-systems-expanding-the-quality-and-efficiency-frontier-of-ai-3/
LOCATION:Amy Gutmann Hall\, Room 414\, 3333 Chestnut Street\, Philadelphia\, 19104\, United States
ORGANIZER;CN="Computer and Information Science":MAILTO:cherylh@cis.upenn.edu
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250404T103000
DTEND;TZID=America/New_York:20250404T114500
DTSTAMP:20260403T160431
CREATED:20250218T204413Z
LAST-MODIFIED:20250218T204413Z
UID:10008294-1743762600-1743767100@seasevents.nmsdev7.com
SUMMARY:Spring 2025 GRASP on Robotics: Reid Simmons\, Carnegie Mellon University\, “AI-Based Assistants for the Elderly”
DESCRIPTION:This will be a hybrid event with in-person attendance in Wu and Chen and virtual attendance on Zoom. \nABSTRACT\nAs the population ages\, the need grows for AI agents to assist people to remain living independently.  Older adults are typically set in their ways\, so AI agents should adapt to their ways of doing things\, rather than the other way around.  To that end\, we are exploring various approaches to learning to personalize assistive agents\, including the use of bandit algorithms\, foundational models\, neuro-symbolic architectures\, and theory of mind.  This talk will present our approaches and results in several assistive areas\, including meal preparation and exercise coaching\, as well as work in learning policies from humans.  Much of the research is being supported by AI-CARING\, an NSF-sponsored Institute devoted to developing AI technologies to help older adults with cognitive and physical decline remain in their homes.
URL:https://seasevents.nmsdev7.com/event/spring-2025-grasp-on-robotics-reid-simmons-carnegie-mellon-university-ai-based-assistants-for-the-elderly/
LOCATION:Wu and Chen Auditorium (Room 101)\, Levine Hall\, 3330 Walnut Street\, Philadelphia\, PA\, 19104\, United States
CATEGORIES:Seminar
ORGANIZER;CN="General Robotics%2C Automation%2C Sensing and Perception (GRASP) Lab":MAILTO:grasplab@seas.upenn.edu
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250404T130000
DTEND;TZID=America/New_York:20250404T150000
DTSTAMP:20260403T160431
CREATED:20250328T182752Z
LAST-MODIFIED:20250328T182752Z
UID:10008338-1743771600-1743778800@seasevents.nmsdev7.com
SUMMARY:Women in Data Science @ Penn Conference: From Data to Discovery: Exploring AI with a Patient Case Study\, ChatGPT and Generative Models
DESCRIPTION:In today’s data-driven landscape\, vast unstructured data sources — like documents and electronic health records (EHRs) — demand advanced AI tools to unlock their full potential. Generative AI\, powered by large language models (LLMs)\, is becoming indispensable for processing and extracting insights from complex language-based data. \nParticipants will explore: \n\nIdentifying key problems that lend themselves to AI-driven solutions.\nCollecting and preparing data\, building models and running state-of-the-art algorithms.\nValidating models and interpreting their results.
URL:https://seasevents.nmsdev7.com/event/women-in-data-science-penn-conference-from-data-to-discovery-exploring-ai-with-a-patient-case-study-chatgpt-and-generative-models/
LOCATION:Jon M. Huntsman Hall\, 3730 Walnut Street\, Philadelphia\, PA\, 19104\, United States
CATEGORIES:AI Month
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250407T140000
DTEND;TZID=America/New_York:20250407T150000
DTSTAMP:20260403T160431
CREATED:20250314T182407Z
LAST-MODIFIED:20250314T182407Z
UID:10008319-1744034400-1744038000@seasevents.nmsdev7.com
SUMMARY:CBE Doctoral Dissertation Defense: "Understanding and Modulating Interactions between Polymers and Nanoparticles for Effective Catalyst Design in Polymer Upcycling" (Anirban Majumder)
DESCRIPTION:Abstract: \n\n\n\nDespite recent advances in catalytic conversion of plastic waste into high-value chemicals\, the interactions between the polymers and catalysts\, which are highly porous nanomaterials\, are not well understood. Fundamental understanding of these interactions and the ability to modulate them would allow us to design effective catalysts for polymer upcycling reactions. To study the interactions between polymers and nanomaterials\, quantifying them is essential. In this thesis\, we employ a quantitative technique to directly measure the contact angle between polymers and nanoparticles\, enabling us to investigate various aspects of polymer-catalyst interactions that are crucial for polymer upcycling reactions. We measure the contact angle of polyolefins with silica nanoparticles\, a commonly used support in heterogeneous catalysis\, and modify the surface chemistry of these nanoparticles by depositing catalytic metals and metal oxides via atomic layer deposition (ALD) and using silane chemistry. Our findings show that the polarizability of polyolefins plays a significant role in their interactions with catalytic support materials\, and modifying the polarity of the support material could be an effective way to tune polymer-catalyst interactions. Further\, we find that the polymer-catalyst interactions are dominated by interactions of the polymers with the support materials and not with the metal catalytic sites. We also probe the influence of common plastic additives\, namely primary antioxidants (PAOs)\, benzophenones and hindered amine light stabilizers (HALS)\, on the polymer-catalyst interactions by adding them to purified high-density polyethylene (HDPE) and measuring the contact angle on silica nanoparticles. We observe that while the addition of PAOs and benzophenones to polyolefins does not affect their interactions with silica significantly\, HALS strongly alter the polymer-silica interactions. Hence\, the presence of different types of additives in plastics might necessitate different strategies to design catalysts for polymer upcycling reactions. Overall\, this thesis sheds light on some key unanswered questions on polymer-catalyst interactions and paves the way for future innovations in polymer upcycling and engineering catalysts for other polymeric reactions.
URL:https://seasevents.nmsdev7.com/event/cbe-doctoral-dissertation-defense-understanding-and-modulating-interactions-between-polyolefins-and-nanoparticles-for-effective-catalyst-design-in-polymer-upcycling-anirban-majumder/
LOCATION:Raisler Lounge (Room 225)\, Towne Building\, 220 South 33rd Street\, Philadelphia\, PA\, 19104\, United States
CATEGORIES:Doctoral,Student,Dissertation or Thesis Defense
ORGANIZER;CN="Chemical and Biomolecular Engineering":MAILTO:cbemail@seas.upenn.edu
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250408T101500
DTEND;TZID=America/New_York:20250408T111500
DTSTAMP:20260403T160431
CREATED:20250324T175817Z
LAST-MODIFIED:20250324T175817Z
UID:10008327-1744107300-1744110900@seasevents.nmsdev7.com
SUMMARY:MEAM Seminar: “Learning Memory and Material Dependent Constitutive Laws”
DESCRIPTION:The theory of homogenization provides a systematic approach to the derivation of macroscale constitutive laws\, obviating the need to repeatedly resolve complex microstructure. However\, the unit cell problem which defines the constitutive model is typically not amenable to analytical solution. It is therefore of interest to learn constitutive models from data generated by the unit cell problem. Many viscoelastic and elastoviscoplastic materials are characterized by memory-dependent constitutive laws. Furthermore\, in order to amortize the computational investment in finding such memory-dependent constitutive laws\, it is desirable to learn their dependence on the material microstructure.\nWhilst the learning of memory dependence and material dependence have been considered separately\, their joint learning has not been considered. This talk is focused on the joint learning problem and proposes a novel neural operator framework to address it \nIn order to provide firm foundations\, the homogenization problem for linear Kelvin–Voigt viscoelastic materials is studied. The theoretical properties of the cell problem\, in this Kelvin–Voigt setting\, are used to motivate the proposed general neural operator framework; these theoretical properties are also used to prove a universal approximation theorem for the learned macroscale constitutive model. This formulation of learnable constitutive models is then deployed beyond the Kelvin–Voigt setting. Numerical experiments are presented showing that the resulting data-driven methodology accurately learns history- and microstructure-dependent linear viscoelastic and nonlinear elastoviscoplastic constitutive models; numerical results also demonstrate that the resulting constitutive models can be deployed in macroscale simulation of material deformation. \nJoint work with Kaushik Bhattacharya\, Lianghao Cao\, George Stepaniants and Margaret Trautner (all Caltech).
URL:https://seasevents.nmsdev7.com/event/meam-seminar-learning-memory-and-material-dependent-constitutive-laws/
LOCATION:Wu and Chen Auditorium (Room 101)\, Levine Hall\, 3330 Walnut Street\, Philadelphia\, PA\, 19104\, United States
CATEGORIES:Seminar
ORGANIZER;CN="Mechanical Engineering and Applied Mechanics":MAILTO:meam@seas.upenn.edu
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250409T120000
DTEND;TZID=America/New_York:20250409T131500
DTSTAMP:20260403T160431
CREATED:20250402T130550Z
LAST-MODIFIED:20250402T130550Z
UID:10008347-1744200000-1744204500@seasevents.nmsdev7.com
SUMMARY:ASSET Seminar: "Alignment and Control with Representation Engineering"
DESCRIPTION:Abstract: \nLarge Language Models (LLMs) are vulnerable to adversarial attacks\, which bypass common safeguards put in place to prevent these models from generating harmful output. Notably\, these attacks can be transferrable to other models—even proprietary ones—potentially compromising a wide range of AI systems with a single exploit. This surprising fragility underscores a critical weakness in current AI safeguards. \nIn this talk\, we illustrate how these attacks are discovered\, and several recent advances that take advantage of models’ internal representations to thwart them. Unlike much prior work that relies on adversarial training methods\, this approach directly controls neural representations responsible for harmful and unwanted behaviors\, while remaining agnostic to particular attacks. Notably\, in start contrast with prior work we show that these methods can remain effective while preserving the model’s performance on non-adversarial inputs. Our findings suggest that achieving robust safety in generative models may be an attainable goal. \nZoom Link:https://upenn.zoom.us/j/95869536469
URL:https://seasevents.nmsdev7.com/event/asset-seminar-alignment-and-control-with-representation-engineering/
LOCATION:Amy Gutmann Hall\, Room 414\, 3333 Chestnut Street\, Philadelphia\, 19104\, United States
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250409T130000
DTEND;TZID=America/New_York:20250409T140000
DTSTAMP:20260403T160431
CREATED:20250403T210257Z
LAST-MODIFIED:20250403T210257Z
UID:10008350-1744203600-1744207200@seasevents.nmsdev7.com
SUMMARY:Spring 2025 GRASP SFI: Lillian Ratliff\, University of Washington\, “Fragile Foundations? Building Robustness into Reasoning with Algorithmic Agents”
DESCRIPTION:This will be a hybrid event with in-person attendance in Levine 307 and virtual attendance on Zoom. \nABSTRACT\nAs AI-enabled systems become integral to critical domains\, their robustness is increasingly tested by dynamic environments\, continual learning\, and inferential uncertainty. Whether an AI proxy informs high-stakes medical decisions or an embodied agent relies on a foundation model to reason across modalities\, today’s training and deployment methodologies remain inherently fragile. This fragility often stems from a reliance on stationarity assumptions\, overly symmetric training paradigms\, and a failure to account for other adapting agents—leading to systems that generalize poorly\, misestimate uncertainty\, and break down in interactive settings.\n\n\nThis talk presents recent theoretical contributions and algorithmic design principles for robust inference and influence when reasoning with algorithmic agents. In particular\, it explores how tools from control and game theory—when integrated into machine learning\, and vice versa—enable uncertainty adaptation and the synthesis of decision-making strategies for influencing algorithmic agents. Through motivating examples\, the talk will illustrate how bridging these disciplines leads to more robust AI systems that can reason\, adapt\, and interact effectively in complex\, non-stationary environments. The first part will focus on algorithms with non-asymptotic convergence guarantees in time-varying settings with a hierarchical game structure. The second part will address uncertainty quantification and adaptation in safety-critical\, multi-agent\, embodied AI systems. The talk will conclude with a discussion of open questions and future directions.
URL:https://seasevents.nmsdev7.com/event/spring-2025-grasp-sfi-lillian-ratliff/
LOCATION:Levine 307\, 3330 Walnut Street\, Philadelphia\, PA\, 19104\, United States
CATEGORIES:Seminar
ORGANIZER;CN="General Robotics%2C Automation%2C Sensing and Perception (GRASP) Lab":MAILTO:grasplab@seas.upenn.edu
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250409T153000
DTEND;TZID=America/New_York:20250409T163000
DTSTAMP:20260403T160431
CREATED:20241216T200401Z
LAST-MODIFIED:20241216T200401Z
UID:10008202-1744212600-1744216200@seasevents.nmsdev7.com
SUMMARY:CBE Seminar: "Value-added Transformations in Electrocatalysis and Graduate Education" (Maureen Tang\, Drexel University)
DESCRIPTION:Abstract: \n\nElectrifying the chemical industry has been much touted as a path to a low-carbon future\, but nearly all pathways of interest are electrochemical reductions. If we want water-to-hydrogen\, CO2-to-chemicals\, or nitrogen-to-ammonia\, from where will we get the electrons? Water-to-oxygen is thermodynamically expensive\, kinetically slow\, and generates a zero-value product. This talk will discuss two potential avenues of investigation for value-added electron-holes. In the first\, we seek to determine the mechanism of six-electron water oxidation to ozone. We discuss our efforts to identify intermediates\, understand the role of dopants\, and deconvolute catalysis from corrosion in electrochemical ozone production. In the second\, we investigate the feasibility of electrocatalytic cyclohexane oxidation. Our results show that molecular oxygen\, not water\, is the primary oxygen source\, with implications for radical intermediates. We furthermore demonstrate the impact of cross-over in these systems and point to the importance of electrochemical reactor design in lab-scale studies. In the final part of this talk\, I will discuss non-technical limitations to the aforementioned approach. I will incorporate advances in behavioral economics and organizational behavior into a novel mentoring activity for PhD students while exploring themes of scarcity and abundance in the modern university.
URL:https://seasevents.nmsdev7.com/event/cbe-seminar-value-added-transformations-in-electrocatalysis-and-graduate-education-maureen-tang-drexel-university/
LOCATION:Wu & Chen Auditorium
CATEGORIES:Seminar
ORGANIZER;CN="Chemical and Biomolecular Engineering":MAILTO:cbemail@seas.upenn.edu
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250410T100000
DTEND;TZID=America/New_York:20250410T110000
DTSTAMP:20260403T160431
CREATED:20250331T173405Z
LAST-MODIFIED:20250331T173405Z
UID:10008343-1744279200-1744282800@seasevents.nmsdev7.com
SUMMARY:MEAM Ph.D. Thesis Defense: "Macroscopic Ensemble Methods for Multi Robot Task Assignment in Dynamic Environments"
DESCRIPTION:With finite resources to complete tasks like monitoring\, coverage\, and search\, the challenge lies in identifying and performing tasks that can change both in frequency and location. One potential solution is teams of robots equipped with the necessary capabilities to complete the desired tasks. However\, robot teams require methods that effectively assign robots to tasks\, which is also known as the Multi Robot Task Allocation (MRTA) problem. Existing MRTA approaches assign each individual robot to a task. This works well if the team is small (less than 20 robots) and if the individual task specification does not change (monitoring a static environment). Unfortunately\, these solutions require solving a combinatorial optimization problem which has poor computational scalability as the team and number of tasks increase\, and is further exacerbated by changing task or environment conditions. An alternative perspective is to look at how biologists study animal resource selection. Instead of posing the problem of assigning animals to resources\, animals randomly select a resource weighted by the perceived resource value resulting in beneficial population configurations. Taking inspiration from biology\, we model team-wide objectives using macroscopic ensemble allocation. These methods allow robots to select stationary task regions\, are known to easily control large robot teams (more than 50 robots)\, and can even describe robot team heterogeneity. Nevertheless\, macroscopic ensemble methods require extensions to effectively solve the MRTA problem for tasks where conditions change\, e.g.\, monitoring spatiotemporal processes. The main contributions of this dissertation include online adaptive macroscopic allocation\, distributed adaptive macroscopic allocation\, and macroscopic allocation via robot-robot collaboration. Our results show robot teams monitoring spatial temporal environments using simulation and robot experiments.
URL:https://seasevents.nmsdev7.com/event/meam-ph-d-thesis-defense-macroscopic-ensemble-methods-for-multi-robot-task-assignment-in-dynamic-environments/
LOCATION:Levine 307\, 3330 Walnut Street\, Philadelphia\, PA\, 19104\, United States
CATEGORIES:Doctoral,Dissertation or Thesis Defense
ORGANIZER;CN="Mechanical Engineering and Applied Mechanics":MAILTO:meam@seas.upenn.edu
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250410T103000
DTEND;TZID=America/New_York:20250410T120000
DTSTAMP:20260403T160431
CREATED:20250322T150920Z
LAST-MODIFIED:20250322T150920Z
UID:10008325-1744281000-1744286400@seasevents.nmsdev7.com
SUMMARY:MSE/MEAM Seminar: "Converting Scientific Discovery and Disruptive Ideas Into Impactful Energy Technologies with ARPA-E" Laurent Pilon: ARPA-E
DESCRIPTION:This talk presents the Advanced Research Projects Agency-Energy (ARPA-E) and how stakeholders at the University of Pennsylvania can engage in a dialog with the funding agency. ARPA-E advances high-risk high-impact transformational technologies to generate\, store\, and use energy. As part of the US Department of Energy\, ARPA-E’s mission is to enhance the economic and energy security of the United States through the development of energy technologies that (i) reduce energy import\, (ii) improve energy efficiency\, (iii) reduce greenhouse gas emissions\, and (iv) ensure the resilience\, reliability\, and security of the U.S. energy infrastructure. Another mission of ARPA-E is to ensure that the United States maintains a technological lead in developing and deploying advanced energy technologies. These missions are achieved through rigorous program development\, competitive project selection\, and active program management to ensure impactful expenditures. We will discuss the different funding mechanisms\, the lifecycle of ARPA-E programs and how stakeholders can engage at different stages. This discussion will be illustrated with the speaker’s experience as a faculty member at UCLA and in his roles at ARPA-E where he developed and managed the following programs: \n\nIGNIITE supporting early-career innovators seeking to convert disruptive ideas into impactful new technologies across the full spectrum of energy applications.\nCIRCULAR developing foundational technologies to achieve a circular electric vehicle battery supply chain.\nSuperconducting Tape program aiming to manufacture high performance\, low cost\, high temperature superconducting tapes to support developments in fusion reactors\, electric aviation\, and power transmission.
URL:https://seasevents.nmsdev7.com/event/mse-seminar-converting-scientific-discovery-and-disruptive-ideas-into-impactful-energy-technologies-with-arpa-e-lauren-pilon-arpa-e/
LOCATION:Wu and Chen Auditorium (Room 101)\, Levine Hall\, 3330 Walnut Street\, Philadelphia\, PA\, 19104\, United States
CATEGORIES:Seminar
ORGANIZER;CN="Materials Science and Engineering":MAILTO:johnruss@seas.upenn.edu
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250410T120000
DTEND;TZID=America/New_York:20250410T131500
DTSTAMP:20260403T160431
CREATED:20250131T221023Z
LAST-MODIFIED:20250131T221023Z
UID:10008261-1744286400-1744290900@seasevents.nmsdev7.com
SUMMARY:IDEAS/STAT Optimization Seminar: "Gradient Equilibrium in Online Learning"
DESCRIPTION:We present a new perspective on online learning that we refer to as gradient equilibrium: a sequence of iterates achieves gradient equilibrium if the average of gradients of losses along the sequence converges to zero. In general\, this condition is not implied by\, nor implies\, sublinear regret. It turns out that gradient equilibrium is achievable by standard online learning methods such as gradient descent and mirror descent with constant step sizes (rather than decaying step sizes\, as is usually required for no regret). Further\, as we show through examples\, gradient equilibrium translates into an interpretable and meaningful property in online prediction problems spanning regression\, classification\, quantile estimation\, and others. Notably\, we show that the gradient equilibrium framework can be used to develop a debiasing scheme for black-box predictions under arbitrary distribution shift\, based on simple post hoc online descent updates. We also show that post hoc gradient updates can be used to calibrate predicted quantiles under distribution shift\, and that the framework leads to unbiased Elo scores for pairwise preference prediction. \n  \n  \nZoom link: https://upenn.zoom.us/j/98220304722
URL:https://seasevents.nmsdev7.com/event/ideas-stat-optimization-seminar-ryan-tibishirani/
LOCATION:Amy Gutmann Hall\, Room 414\, 3333 Chestnut Street\, Philadelphia\, 19104\, United States
CATEGORIES:Seminar,Colloquium
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250411T103000
DTEND;TZID=America/New_York:20250411T114500
DTSTAMP:20260403T160431
CREATED:20250331T195717Z
LAST-MODIFIED:20250331T195717Z
UID:10008344-1744367400-1744371900@seasevents.nmsdev7.com
SUMMARY:Spring 2025 GRASP on Robotics: Lingjie Liu\, University of Pennsylvania\, “Towards Next-Gen 3D Reconstruction and Generation: From Visual Fidelity to Multimodal and Physical Understanding”
DESCRIPTION:This will be a hybrid event with in-person attendance in Wu and Chen and virtual attendance on Zoom. This seminar will NOT be recorded. \nABSTRACT\nRecent years have witnessed remarkable progress in 3D reconstruction and generation. However\, most existing methods primarily focus on modeling geometry and appearance. I believe the next generation of 3D reconstruction and generation should go further in two key directions. First\, it should be well-aligned with other modalities—such as language and images—so that 3D representations can play an important role in the multi-modal era. Second\, it should incorporate physical understanding to ensure reconstructions and generations are physically plausible\, which will ultimately make them more applicable in robotics. In this talk\, I will present our recent efforts toward these goals and discuss the challenges that lie ahead.
URL:https://seasevents.nmsdev7.com/event/spring-2025-grasp-on-robotics-lingjie-liu-university-of-pennsylvania-towards-next-gen-3d-reconstruction-and-generation-from-visual-fidelity-to-multimodal-and-physical-understanding/
LOCATION:Wu and Chen Auditorium (Room 101)\, Levine Hall\, 3330 Walnut Street\, Philadelphia\, PA\, 19104\, United States
CATEGORIES:Seminar
ORGANIZER;CN="General Robotics%2C Automation%2C Sensing and Perception (GRASP) Lab":MAILTO:grasplab@seas.upenn.edu
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250411T113000
DTEND;TZID=America/New_York:20250411T130000
DTSTAMP:20260403T160431
CREATED:20250408T153812Z
LAST-MODIFIED:20250408T153812Z
UID:10008355-1744371000-1744376400@seasevents.nmsdev7.com
SUMMARY:CBE Doctoral Dissertation Defense: "Context-Dependent Protein Surface Hydrophilicity" (Lilia F. Escobedo)
DESCRIPTION:Abstract: \nProteins perform many important biological functions while also avoiding fouling in an aqueous and crowded cellular environment. Their surfaces have evolved to be both chemically heterogeneous (containing nonpolar\, polar\, and charged groups) as well as hydrophilic. While nonpolar groups are known to induce hydrophobicity\, surface heterogeneity has been found to enhance hydrophilicity and the resistance of non-specific adsorption of biomolecules. Yet\, the exact relationship between the nature of heterogeneous surfaces and hydrophilicity is not fully understood. Therefore\, protein surfaces offer a promising avenue of study to help elucidate this relationship. While many characterizations of protein surface hydrophilicity sum the individual hydrophilicity of amino acids\, surface hydration of heterogeneous surfaces has been found to be highly context-dependent and thus non-additive. In this dissertation\, we utilize molecular dynamics simulations to characterize the atomic-level\, context-dependent hydrophilicity of protein surfaces to better understand how the chemical composition and surface patterning of protein surfaces enhance hydrophilicity. We demonstrate that charged moieties play a much more significant role in enhancing protein surface hydrophilicity than polar atoms do. In fact\, we also demonstrate that chemical composition alone is insufficient to distinguish between hydrophilic and hydrophobic protein surface regions. Furthermore\, we use these findings to develop protein-inspired design rules for heterogeneous non-fouling surfaces. The work in this dissertation could be used to inform the design of superhydrophilic materials as well as to elucidate the relationship between surface heterogeneity and hydrophilicity. Additionally\, it could be used to not only better understand biomolecular interactions through a context-dependent characterization of protein hydrophilicity\, but also to inform protein engineering. \nZoom Information:\nMeeting ID: 953 9917 7738\nPasscode: 805166
URL:https://seasevents.nmsdev7.com/event/cbe-doctoral-dissertation-defense-context-dependent-protein-surface-hydrophilicity-lilia-f-escobedo/
LOCATION:Vagelos Institute for Energy Science and Technology\, Room 121\, 231 S 34th Street\, Philadelphia\, PA\, 19104\, United States
CATEGORIES:Doctoral,Student,Dissertation or Thesis Defense
ORGANIZER;CN="Chemical and Biomolecular Engineering":MAILTO:cbemail@seas.upenn.edu
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250411T130000
DTEND;TZID=America/New_York:20250411T141500
DTSTAMP:20260403T160431
CREATED:20250328T183142Z
LAST-MODIFIED:20250328T183142Z
UID:10008339-1744376400-1744380900@seasevents.nmsdev7.com
SUMMARY:Penn Engineering Entrepreneurship (EENT): Generative AI Panel
DESCRIPTION:A panel hosted by Penn Engineering Entrepreneurship (EENT) highlighting the importance of AI for the next generation of leaders and showcasing how Penn has been\, and will continue to be\, at the forefront of this evolving field. \nPanelists include: \nElizabeth (Liz) Golden\nCEO & Co-Founder @ Wavelet Medical \nMel Tang\nFounding Operating Partner & CFO of Matter Venture Partners; Formerly CFO for Ring.com \nNat Trask\nAssociate Professor of Mechanical Engineering and Applied Mechanics; formerly Principal @ Sandia National Laboratories \nMark Weber\nDirector’s Fellow of MIT Media Lab & Investor; formerly Strategy and Ops Lead @ MIT-IBM Watson AI Lab \nBrian Halak\nPractice Professor of Engineering Entrepreneurship and Managing Partner @ Medical Excellence Capital
URL:https://seasevents.nmsdev7.com/event/penn-engineering-entrepreneurship-eent-generative-ai-panel/
LOCATION:Amy Gutmann Hall\, Auditorium\, 3333 Chestnut Street\, Philadelphia\, PA\, 19104\, United States
CATEGORIES:Conference,AI Month
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250411T140000
DTEND;TZID=America/New_York:20250411T150000
DTSTAMP:20260403T160431
CREATED:20250130T153134Z
LAST-MODIFIED:20250130T153134Z
UID:10008249-1744380000-1744383600@seasevents.nmsdev7.com
SUMMARY:PICS Colloquium: Multiscale simulations of soft matter: from block copolymers to biomolecular condensates
DESCRIPTION:Polymers are ubiquitous in both synthetic and biological materials and underlie technologies as diverse as surfactants\, adhesives\, proteins and DNA. One of the defining features of all polymeric materials is that they are characterized by a wide range of length scales\, often involving phenomena that span nanometers to microns. This hierarchy of length scales presents significant challenges for polymeric simulations. Existing models that can resolve phenomena at monomeric length scales (e.g. Atomistic simulations) are far too expensive to access mesoscopic processes\, whereas coarse-grained models that can access these large length scales must necessarily omit chemical details\, which often enter calculations through phenomenological parameters. As such\, there is a significant need for new computational techniques that can accurately and efficiently predict how chemical changes at the smallest length scales will propagate up to the largest length scales in a material. \n  \nIn this talk\, I describe recent efforts by my group to develop multiscale simulations of polymeric materials that can link monomeric to mesoscopic length scales. In the first part of the talk\, I describe a new “multi-representation” simulation method where particle-based and field-theoretic simulations are linked together into a unified framework. This approach can accelerate polymer simulations by several orders of magnitude and can rapidly equilibrate mesoscopic length scales while preserving monomer-scale details and dynamics. Notably\, this multi-representation approach leverages the formal equivalence between particle and field-based models and involves no approximation. The utility of this approach is illustrated by examining the self-assembly of complex sphere phases in block copolymer melts. In the second part of the talk\, I demonstrate how these multi-representation simulations can be extended to explain the phase separation of biomolecular condensates. Our approach can model proteins at amino acid resolution yet is efficient enough to access the long length and time scales that characterize condensate formation. Our approach can recapitulate recent experimental data on prion-like domains and can resolve how subtle modifications to the amino acid sequence can modulate their phase behavior. Finally\, we examine the phase separation of chromatin and how post-translational modifications to histone proteins can lead to transitions between liquid-like and solid-like behavior. Taken together\, this work demonstrates that multi-representation simulations that combine particle and field-theoretic simulations can unlock new insights into the multiscale physics that characterize synthetic and biological polymers.
URL:https://seasevents.nmsdev7.com/event/pics-colloquium-with-josh-lequieu/
LOCATION:PICS Conference Room 534 – A Wing \, 5th Floor\, 3401 Walnut Street\, Philadelphia\, PA\, 19104\, United States
CATEGORIES:Colloquium
ATTACH;FMTTYPE=image/jpeg:https://seasevents.nmsdev7.com/wp-content/uploads/2025/01/lequieu.jpg
ORGANIZER;CN="Penn Institute for Computational Science (PICS)":MAILTO:dkparks@seas.upenn.edu
END:VEVENT
BEGIN:VEVENT
DTSTART;VALUE=DATE:20250414
DTEND;VALUE=DATE:20250416
DTSTAMP:20260403T160431
CREATED:20250410T192829Z
LAST-MODIFIED:20250410T192829Z
UID:10008359-1744588800-1744761599@seasevents.nmsdev7.com
SUMMARY:AI Infrastructure: Foundations for Energy Efficiency and Scalability
DESCRIPTION:Click here for more details. \nThe workshop will explore the state of the art in sustainable computing and share recent research at the intersection of technology\, economics\, and policy. Through invited talks\, panel discussions\, and breakout sessions\, participants will help shape a research agenda for the field. The workshop aims to produce a white paper and publish a research perspective in a major journal or media outlet. \nOrganization. The workshop is organized by Carbon Connect (https://carbonconnect.eco/)\, a team of researchers who have been awarded an Expedition in Computing by the National Science Foundation. It is co-hosted by Penn Engineering and Wharton School. \nThe workshop program agenda is available here.
URL:https://seasevents.nmsdev7.com/event/ai-infrastructure-foundations-for-energy-efficiency-and-scalability/
LOCATION:Jon M. Huntsman Hall\, 3730 Walnut Street\, Philadelphia\, PA\, 19104\, United States
CATEGORIES:Conference,AI Month
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250414T150000
DTEND;TZID=America/New_York:20250414T160000
DTSTAMP:20260403T160431
CREATED:20250409T190428Z
LAST-MODIFIED:20250409T190428Z
UID:10008356-1744642800-1744646400@seasevents.nmsdev7.com
SUMMARY:Spring 2025 GRASP Seminar: Sebastian Scherer\, Carnegie Mellon University\, "Resilient Autonomy for Extreme and Uncertain Environments"
DESCRIPTION:This will be an IN-PERSON event ONLY with in-person attendance in AGH 306. \nABSTRACT\nRobots show great promise if they can get out of the lab into the field and go beyond a single-operator per robot paradigm. However\, the unstructured nature of the real-world requires nuanced decision making of the robot. In this talk I will outline some of our approaches\, progress\, and results on multi-modal sensing\, providing nuanced perception inputs\, as well as navigation in difficult terrain\, and extensions to multi-robot teams\, and future directions of our research.
URL:https://seasevents.nmsdev7.com/event/spring-2025-grasp-seminar-sebastian-scherer-carnegie-mellon-university-resilient-autonomy-for-extreme-and-uncertain-environments/
LOCATION:Amy Gutmann Hall\, Room 306\, 3317 Chestnut Street\, Philadelphia\, PA\, 19104\, United States
CATEGORIES:Seminar
ORGANIZER;CN="General Robotics%2C Automation%2C Sensing and Perception (GRASP) Lab":MAILTO:grasplab@seas.upenn.edu
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250415T101500
DTEND;TZID=America/New_York:20250415T111500
DTSTAMP:20260403T160431
CREATED:20241126T202355Z
LAST-MODIFIED:20241126T202355Z
UID:10008187-1744712100-1744715700@seasevents.nmsdev7.com
SUMMARY:MEAM Seminar: "Microscopic Mayhem: Cancer in Three-Dimensions"
DESCRIPTION:Critical to advancing immunotherapy and cell therapy in cancer is developing a deeper understanding pf the dynamics of immune cell-mediated cytotoxicity. The results from the multidisciplinary effort reported here include numerous measurements and movies of immune cell-mediated cytotoxicity with striking examples of serial killing\, foraging\, path-tracking\, cytokine gradients at tumor margins\, and killing dynamics\, in some cases revealing peak apoptotic signatures just minutes after T Cell engagement. \nIn vitro studies of immune cell killing are traditionally performed using time-lapse imaging and biochemical assays\, but these methods are often limited by spatial and temporal resolution\, throughput\, and the ability to extract the dynamics of cellular interactions. This study integrates high-resolution and high-speed laser scanning confocal microscopy with artificial intelligence (AI)\, and machine learning (ML) approaches to provide a high-resolution data-driven analysis of immune cell killing dynamics in vitro. \nWe have engineered a perfusion-enabled 3D culture system integrated microscopy to assess cellular dynamics for extended periods of time. Perfusion culture maintains the interstitial flow of liquid culture media\, clearing the microenvironment of toxic metabolites and reactive oxygen species. This platform uses a Liquid-Like Solids (LLS) to mimic the transport dynamics of a capillary bed. Integrated microscopy allows in situ quantification of spatiotemporal cytokine concentrations\, immune cell tracking\, immune cell killing dynamics\, and invasion dynamics.
URL:https://seasevents.nmsdev7.com/event/meam-seminar-microscopic-mayhem-cancer-in-three-dimensions/
LOCATION:Wu and Chen Auditorium (Room 101)\, Levine Hall\, 3330 Walnut Street\, Philadelphia\, PA\, 19104\, United States
CATEGORIES:Seminar
ORGANIZER;CN="Mechanical Engineering and Applied Mechanics":MAILTO:meam@seas.upenn.edu
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250415T120000
DTEND;TZID=America/New_York:20250415T133000
DTSTAMP:20260403T160431
CREATED:20250328T185355Z
LAST-MODIFIED:20250328T185355Z
UID:10008340-1744718400-1744723800@seasevents.nmsdev7.com
SUMMARY:What Does AI Tell Us About What It Means to Be Human
DESCRIPTION:Please RSVP here. \nWe are living in an age where capabilities previously thought to be hallmarks of human intelligence are increasingly being replicated\, or at least mimicked\, in artificial systems. \nTasks involving language\, reasoning\, perception and even interaction with the real world have all been demonstrated in silico. What does this fact tell us about the nature of human intelligence? In what ways is human intelligence similar to or different from machine intelligence? How have philosophers and other thinkers conceptualized issues related to thought and agency over the centuries and which of these ideas are relevant to today’s context? \nJoin us for a panel discussion where faculty members from Philosophy\, Cognitive Science and Computer Science will exchange perspectives on these issues. \nCamillo Jose Taylor (Moderator)\nRaymond S. Markowitz President’s Distinguished Professor\nComputer and Information Science \nChris Callison Burch\nProfessor of Computer and Information Science\nProgram Director of Online Master of Science in Engineering in Artificial Intelligence \nCharles Yang\nProfessor of Linguistics and Computer Science\nDirector\, Program in Cognitive Science \nCarlos Gray Santana\nAssociate Professor of Philosophy
URL:https://seasevents.nmsdev7.com/event/what-does-ai-tell-us-about-what-it-means-to-be-human/
LOCATION:Wu and Chen Auditorium (Room 101)\, Levine Hall\, 3330 Walnut Street\, Philadelphia\, PA\, 19104\, United States
CATEGORIES:AI Month
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250415T140000
DTEND;TZID=America/New_York:20250415T160000
DTSTAMP:20260403T160431
CREATED:20250328T185934Z
LAST-MODIFIED:20250328T185934Z
UID:10008341-1744725600-1744732800@seasevents.nmsdev7.com
SUMMARY:AI Month Alumni Panel
DESCRIPTION:Penn Engineering alumni working in AI will share industry insights in a panel discussion\, followed by student networking sessions. \nSchedule: \n2–3 p.m.  Panel Discussion and Q&A\n3:15–4 p.m.  Breakout Networking Sessions \nPanelists include: \nSara Dwyer (ENG’19)\nFounder & CEO at Parambil \nDavid Q. Sun (GEE’14\, GR’20)\nSenior Engineering Manager\, Siri & Information Intelligence\, AIML at Apple \nArchana Vemulapalli (GEN’01)\nCorporate Vice President\, Global Commercial Sales at AMD \nModerated by\nGeorge Pappas\nUPS Foundation Professor of Transportation in Computer and Information Science\, in Electrical and Systems Engineering and in Mechanical Engineering and Applied Mechanics
URL:https://seasevents.nmsdev7.com/event/ai-month-alumni-panel/
LOCATION:Berger Auditorium (Room 13)\, Skirkanich Hall\, 210 South 33rd Street\, Philadelphia\, PA\, 19104\, United States
CATEGORIES:Alumni,AI Month
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250415T153000
DTEND;TZID=America/New_York:20250415T173000
DTSTAMP:20260403T160431
CREATED:20250403T205157Z
LAST-MODIFIED:20250403T205157Z
UID:10008349-1744731000-1744738200@seasevents.nmsdev7.com
SUMMARY:ESE Ph.D. Thesis Defense: ''Manifold Filters and Neural Networks: Geometric Graph Signal Processing in the Limit''
DESCRIPTION:Graph Neural Networks (GNNs) are the tool of choice for scalable and stable learning in graph-structured data applications involving geometric information. My research addresses the fundamental questions of how GNNs can generalize across different graph scales and how they can remain stable on large-scale graphs. I do so by considering manifolds as graph limit models. In this talk\, we will explain how to build manifold convolutional filters and manifold neural networks (MNNs) as the limit objects of graph convolutional filters and GNNs when the graphs are sampled from manifolds. Using the Laplace-Beltrami operator exponentials to define manifold convolutions\, we demonstrate their algebraic equivalence to both graph convolutions and standard time convolutions in nodal and spectral domains. This equivalence provides a unifying framework to analyze key theoretical properties of GNNs: i) Convergence of GNNs to MNNs allows the scalability of GNNs on graphs across scales. ii) The stability of MNNs to deformations indicates the stability of large-scale GNNs. These findings offer practical guidelines for designing GNN architectures\, particularly by imposing constraints on the spectral properties of filter functions. Theoretical results are verified in real-world scenarios\, including point cloud analysis\, wireless resource allocation\, and wind field studies on vector fields.
URL:https://seasevents.nmsdev7.com/event/ese-ph-d-thesis-defense-manifold-filters-and-neural-networks-geometric-graph-signal-processing-in-the-limit/
LOCATION:Amy Gutmann Hall\, Room 515\, 3317 Chestnut Street\, Philadelphia\, 19104\, United States
CATEGORIES:Dissertation or Thesis Defense
ORGANIZER;CN="Electrical and Systems Engineering":MAILTO:eseevents@seas.upenn.edu
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250416T120000
DTEND;TZID=America/New_York:20250416T131500
DTSTAMP:20260403T160431
CREATED:20250404T165546Z
LAST-MODIFIED:20250404T165546Z
UID:10008352-1744804800-1744809300@seasevents.nmsdev7.com
SUMMARY:ASSET Seminar: "Learning Reliable and Robust Generative Intelligence"
DESCRIPTION:Abstract: \nRobust simulation and precise modeling of physical dynamics are essential for advancing perception\, planning\, and control in the development of generalist physical agents. In this talk\, I will present my research on building generative models that combine physical realism with scalability in high-dimensional environments. The presentation delves into both the theoretical foundations and practical implementations of our methods\, including the incorporation of 3D constraints into video diffusion models and the integration of autoregressive structures into continuous generative modeling to better handle complex data. By combining these techniques\, the models replicate intricate interactions in dynamic scenes and demonstrate their potential to support efficient\, data-driven learning across a broad range of applications. \nZoom Link: https://upenn.zoom.us/j/92594955973 \n 
URL:https://seasevents.nmsdev7.com/event/asset-seminar-jiatao-gu-university-of-pennsylvania/
LOCATION:Amy Gutmann Hall\, Room 414\, 3333 Chestnut Street\, Philadelphia\, 19104\, United States
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250416T143000
DTEND;TZID=America/New_York:20250416T143000
DTSTAMP:20260403T160431
CREATED:20250411T143317Z
LAST-MODIFIED:20250411T143317Z
UID:10008360-1744813800-1744813800@seasevents.nmsdev7.com
SUMMARY:ESE Ph.D. Thesis Defense: "Training Adaptive and Sample-Efficient Autonomous Agents"
DESCRIPTION:AI agents\, both in the physical and digital worlds\, should generalize from their training data to three increasingly difficult levels of deployment: training tasks and environments\, training tasks and environments with variations\, and completely new tasks and environments. Moreover\, like humans\, they are expected to learn from as little training data as possible\, especially in the physical world\, and adapt with as little adaptation data as possible. This thesis is founded around and describes work that tackles these levels of generalization with an additional emphasis on sample-efficiency. \nWe start with a focus on training data efficiency and the simplest level of generalization from training data to training tasks and environments (a.k.a.\, level 1). AI agents\, especially in the physical world\, are usually trained via one of two paradigms: imitation learning or reinforcement learning. First\, we propose a plug-in model class to improve behavior cloning with any deep neural network (DNN) backbone that is particularly effective in the low-data regime. Second\, we leverage our proposed model class to guarantee the conformance of any DNN world model to physics and medical constraints\, in a highly data-efficient manner. Third\, we improve the sample-efficiency of reinforcement learning agents\, by an order of magnitude\, by leveraging expert interventions. \nNext\, we tackle the challenge of generalization to training tasks and environments with variations as well as completely new tasks and environments (a.k.a.\, levels 2 and 3)\, keeping both training and adaptation sample-efficiency in mind. Here\, we pre-train REGENT\, a retrieval-augmented generalist agent that can adapt to unseen robotics and game-playing environments via in-context learning\, without any finetuning. REGENT outperforms state-of-the-art generalist agents after pre-training on an order-of-magnitude fewer datapoints and with up to 3x fewer parameters. We also propose a strategy\, inspired by adaptive control\, to improve the robustness of the image encoder of REGENT\, an essential component for handling environment variations. \nFinally\, we bring REGENT to the real world by converting a Vision Language Action model (VLA) to a REGENTic VLA capable of generalizing to unseen objects and tasks through retrieval-augmentation and in-context learning. Further task-specific REGENTic-tuning substantially improves reliability\, surpassing a VLA directly fine-tuned on the same data. \nWe conclude by outlining future directions to expand the envelope of tasks and environments to which a general AI agent can adapt.
URL:https://seasevents.nmsdev7.com/event/ese-ph-d-thesis-defense-training-adaptive-and-sample-efficient-autonomous-agents/
LOCATION:Room 512\, Levine Hall\, 3330 Walnut Street\, Philadelphia\, PA\, 19104\, United States
CATEGORIES:Dissertation or Thesis Defense
ORGANIZER;CN="Electrical and Systems Engineering":MAILTO:eseevents@seas.upenn.edu
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250416T150000
DTEND;TZID=America/New_York:20250416T160000
DTSTAMP:20260403T160431
CREATED:20250411T145839Z
LAST-MODIFIED:20250411T145839Z
UID:10008361-1744815600-1744819200@seasevents.nmsdev7.com
SUMMARY:Spring 2025 GRASP SFI: Anastasia Bizyaeva\, Cornell University\, “Nonlinear dynamics of social decision-making and belief formation”
DESCRIPTION:This will be a hybrid event with in-person attendance in Levine 307 and virtual attendance on Zoom. \nABSTRACT\nMotivated by the study of complex social behavior and by the bottom-up design of collaborative autonomy\, we present and analyze a nonlinear dynamic model of social belief formation. In our framework\, belief updates of individuals are informed by the interplay of external factors\, i.e. social network effects\, and internal factors\, i.e. individual biases and interdependencies between different belief dimensions. The model accounts for networked relationships between an individual’s internal belief representations and nonlinear processing of social information. Our analysis sheds light on mechanistic principles that enable groups to make fast and flexible collective decisions\, overcoming deadlocks to form strong beliefs when it is urgent to do so. We show how the structure of a social network and of the underlying belief system graph shapes emergent social decisions in the group\, and how effects of local biases and inputs can be amplified through feedback to enable tunably sensitive informed collective decisions. This work provides novel insights into the dynamics of complex social systems and motivates a new approach for the design of distributed decision-making strategies in engineered networks of social agents.
URL:https://seasevents.nmsdev7.com/event/spring-2025-grasp-sfi-anastasia-bizyaeva/
LOCATION:Levine 307\, 3330 Walnut Street\, Philadelphia\, PA\, 19104\, United States
CATEGORIES:Seminar
ORGANIZER;CN="General Robotics%2C Automation%2C Sensing and Perception (GRASP) Lab":MAILTO:grasplab@seas.upenn.edu
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250416T153000
DTEND;TZID=America/New_York:20250416T163000
DTSTAMP:20260403T160431
CREATED:20241223T152048Z
LAST-MODIFIED:20241223T152048Z
UID:10008205-1744817400-1744821000@seasevents.nmsdev7.com
SUMMARY:CBE Seminar: "Computational Design and Simulations of Soft Matter: From Molecular Insights to Functional Materials" (Antonia Statt\, UIUC)
DESCRIPTION:Abstract: \nI will present the phase separation behavior of different sequences of a coarse-grained model for sequence defined macromolecules. They exhibit a surprisingly rich phase behavior\, and not only conventional liquid-liquid phase separation is observed\, but also reentrant phase behavior. Most sequences form open phases consisting of large\, interconnected aggregates (e.g. string-like or membrane-like clusters)\, rather than a conventional dense liquids. Minor alterations in the sequence may lead to large changes in the overall phase behavior\, a fact of significant potential relevance for biology and for designing self-assembled structures using block copolymers. I will also discuss recent results from unsupervised manifold learning (UMAP) to classify the different aggregate types and what we can learn from machine learning. Using a bidirectional Recurrent Neural Network (RNN)\, we can now predict which sequence will self-assemble into what aggregate structure. With this framework\, we can investigate the effects of dispersity and sequence errors\, which is of immediate importance for experimental investigations. \nAdditionally\, I will briefly discuss how block copolymers interact with biological lipid membranes to form hybrid membranes. Hybrid phospholipid block copolymer bilayers display many properties\, seen in biomembranes such as selective transport phenomena\, synergistic elastic properties\, and structural phase transformations. Just like in biomembranes\, these fundamental properties of hybrid bilayers are often regulated by lateral phase separation. Understanding the molecular and physical cues that determine the formation of rafts or domains in hybrid membranes\, their size\, and morphology is paramount to elucidating and programming their function. We find that at low polymer content\, a new structure develops in which the bilayer leaflets unzip (but remain continuous) to incorporate nanodomains of hydrophobic butadiene globules. Our findings offer new insights into the morphology of biomembranes upon the insertion of transmembrane proteins with bulky hydrophobic residues.
URL:https://seasevents.nmsdev7.com/event/cbe-seminar-computational-design-and-simulations-of-soft-matter-from-molecular-insights-to-functional-materials-antonia-statt-uiuc/
LOCATION:Wu & Chen Auditorium
CATEGORIES:Seminar
ORGANIZER;CN="Chemical and Biomolecular Engineering":MAILTO:cbemail@seas.upenn.edu
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250417T103000
DTEND;TZID=America/New_York:20250417T120000
DTSTAMP:20260403T160431
CREATED:20250219T155903Z
LAST-MODIFIED:20250219T155903Z
UID:10008296-1744885800-1744891200@seasevents.nmsdev7.com
SUMMARY:MSE Seminar: "Foundry Enabled Chip-Scale Photonics Technology and Applications"   Shaya Fainman - University of California\, San Diego (UCSD)
DESCRIPTION:Dense photonic integration requires miniaturization of materials\, devices\, circuits and systems\, including passive components (e.g.\, engineered composite metamaterials\, filters\, etc.)\, active components (e.g.\, modulators and nonlinear wave mixers) and integrated circuits (Fourier transform spectrometer\, programmable phase modulator of free space modes\, linear algebra processors\, etc.). In this talk we will discuss recent progress in developing CMOS compatible nonlinear optical materials as well as examples of foundry enabled silicon photonic circuits and systems. Specifically\, we will review silicon photonics-based Fourier transform spectrometer (Si-FTS) that can bring broadband operation and fine resolution to the chip scale. Here we will present the modeling and experimental demonstration of a thermally tuned Si-FTS accounting for dispersion\, thermo-optic non-linearity\, and thermal expansion.  We show how these effects modify the relation between the spectrum and interferogram of a light source and we develop a quantitative correction procedure through calibration with a tunable laser. Providing design flexibility and robustness\, the Si-FTS is poised to become a fundamental building block for on-chip spectroscopy. Moreover\, taking advantage of nanofabrication we will discuss on-chip spectrometers using stratified waveguide filters and machine learning. Moving forward\, we will discuss chip-scale integrated circuit/system that will allow to realize linear algebra accelerators with superior performance in speed\, energy consumption and size compared to its electronic counterpart. Such a system can be manufactured using monolithic CMOS process and impact such applications as 5G/6G and beyond wireless MIMO systems as well as deep learning and artificial intelligence.
URL:https://seasevents.nmsdev7.com/event/mse-seminar-foundry-enabled-chip-scale-photonics-technology-and-applications-shaya-fainman-university-of-california-san-diego-ucsd/
LOCATION:Wu and Chen Auditorium (Room 101)\, Levine Hall\, 3330 Walnut Street\, Philadelphia\, PA\, 19104\, United States
CATEGORIES:Seminar
ORGANIZER;CN="Materials Science and Engineering":MAILTO:johnruss@seas.upenn.edu
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250417T120000
DTEND;TZID=America/New_York:20250417T131500
DTSTAMP:20260403T160431
CREATED:20250131T222300Z
LAST-MODIFIED:20250131T222300Z
UID:10008262-1744891200-1744895700@seasevents.nmsdev7.com
SUMMARY:IDEAS/STAT Optimization Seminar: Resilient Distributed Optimization for Cyberphysical Systems
DESCRIPTION:Zoom link: https://upenn.zoom.us/j/98220304722 \n  \nAbstract:\nThis talk considers the problem of resilient distributed multi-agent optimization for cyberphysical systems in the presence of malicious or non-cooperative agents. It is assumed that stochastic values of trust between agents are available which allows agents to learn their trustworthy neighbors simultaneously with performing updates to minimize their own local objective functions. The development of this trustworthy computational model combines the tools from statistical learning and distributed consensus-based optimization. Specifically\, we derive a unified mathematical framework to characterize convergence\, deviation of the consensus from the true consensus value\, and expected convergence rate\, when there exists additional information of trust between agents. Under certain conditions\, we show  that the consensus protocol has almost sure convergence to a common limit value is possible even when malicious agents constitute more than half of the network;  the deviation of the converged limit\, from the nominal no attack case can be bounded with probability that approaches 1 exponentially\, and that correct classification of malicious and legitimate agents can be attained in (random) finite time almost surely. Further\, the expected convergence rate decays exponentially with the quality of the trust observations between agents. We then combine the trust-learning model within a distributed gradient-based method for solving a multi-agent optimization problem and characterize its performance.
URL:https://seasevents.nmsdev7.com/event/ideas-stat-optimization-seminar-angelia-nedich/
LOCATION:Amy Gutmann Hall\, Room 414\, 3333 Chestnut Street\, Philadelphia\, 19104\, United States
CATEGORIES:Seminar,Colloquium
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250417T153000
DTEND;TZID=America/New_York:20250417T163000
DTSTAMP:20260403T160431
CREATED:20250410T191631Z
LAST-MODIFIED:20250410T191631Z
UID:10008358-1744903800-1744907400@seasevents.nmsdev7.com
SUMMARY:CIS Seminar: "Correctness Matters: Automatic Software Engineering in the age of Generative AI"
DESCRIPTION:Software engineers never start from a blank page\, but rather from an extant and usually long-running project in need of modification (for repair\, extension\, update\, etc.). One way to view modern programming is thus as a continual process of iteratively transforming existing programs into something new\, and hopefully better. \nIn this talk\, I will discuss my work on techniques to automate a broad range of software engineering and programming tasks. I position program transformation as a search problem over a space of potential program edits; automating transformation entails careful design choices to manage and successfully traverse this trivially infinite space.  I will focus especially on the fundamental challenge of ensuring that automatically transformed code is of acceptable quality\, and ways to tackle that challenge\, especially in light of recent advances in generative AI. Throughout\, I will highlight my vision of how to develop future-generation tools to help engineers make better software\, and make existing software better\, by carefully integrating domain knowledge and semantics-based reasoning with powerful heuristic search.
URL:https://seasevents.nmsdev7.com/event/cis-seminar-correctness-matters-automatic-software-engineering-in-the-age-of-generative-ai/
LOCATION:Levine 307\, 3330 Walnut Street\, Philadelphia\, PA\, 19104\, United States
ORGANIZER;CN="Computer and Information Science":MAILTO:cherylh@cis.upenn.edu
END:VEVENT
BEGIN:VEVENT
DTSTART;TZID=America/New_York:20250417T173000
DTEND;TZID=America/New_York:20250417T190000
DTSTAMP:20260403T160431
CREATED:20250321T175357Z
LAST-MODIFIED:20250321T175357Z
UID:10008323-1744911000-1744916400@seasevents.nmsdev7.com
SUMMARY:The Future of AI: A Fireside Chat with Yann LeCun\, Chief AI Scientist at Meta
DESCRIPTION:
URL:https://seasevents.nmsdev7.com/event/the-future-of-ai-a-fireside-chat-with-yann-lecun-chief-ai-scientist-at-meta/
LOCATION:Amy Gutmann Hall\, Auditorium\, 3333 Chestnut Street\, Philadelphia\, PA\, 19104\, United States
CATEGORIES:AI Month
END:VEVENT
END:VCALENDAR