Select a calendar:
Filter March Events by Event Type:
SUNMONTUEWEDTHUFRISAT
Conferences, Lectures, & Seminars
Events for March
-
CS Colloquium: Alexis E. Block (UCLA) - Towards Enhanced Social-Physical Human-Robot Interaction
Wed, Mar 01, 2023 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Alexis E. Block, UCLA
Talk Title: Towards Enhanced Social-Physical Human-Robot Interaction
Series: CS Colloquium
Abstract: Hugs are one of the first forms of contact and affection humans experience. Receiving a hug is one of the best ways to feel socially supported, and the lack of social touch can have severe adverse effects on an individual's well-being. Due to the prevalence and health benefits of hugging, we were interested in creating robots that can hug humans as seamlessly as humans hug other humans. However, hugs are complex affective interactions that need to adapt to the height, body shape, and preferences of the hugging partner, and they often include intra-hug gestures like squeezes. In this talk, I'll present the eleven design guidelines of natural and enjoyable robotic hugging that informed the creation of a series of hugging robots that use visual and haptic perception to provide enjoyable interactive hugs. Then, I'll share how each of the four presented HuggieBot versions is evaluated by measuring how users emotionally and behaviorally respond to hugging it. Next, I'll briefly touch on how HuggieBot 4.0 is explicitly compared to a human hugging partner using physiological measures. Finally, I'll share some other forms of physical human-robot interaction I've been working on during my post doc as well as future directions of my research in the area of social-physical human-robot interaction.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Alexis E. Block is currently a postdoctoral research fellow at the University of California, Los Angeles (UCLA), where she is funded by a competitive postdoctoral Computing Innovation Fellowship (CI Fellows) from the US National Science Foundation. She received her Bachelor's in Mechanical Engineering and Applied Science from the University of Pennsylvania in 2016, and her Master's in Robotics in 2017, also from Penn. Block received her Dr. Sc. in Computer Science from ETH Zürich in August 2021, as part of the Max Planck ETH Center for Learning Systems, supervised by Katherine Kuchenbecker, Otmar Hilliges, and Roger Gassert. She was awarded an Otto Hahn Medal from the Max Planck Society for her doctoral work and the Best Hands-On Demonstration at EuroHaptics 2022. Block is currently the General Chair for the Robotics Gordon Research Seminar 2024 and organized the 2022 Southern California Robotics Symposium that took place in September. Alexis's research has been featured in the New York Times, The Times, IEEE Spectrum (twice), NPR, and Nature Outlook to name a few.
Host: Heather Culbertson
Location: Ronald Tutor Hall of Engineering (RTH) - 109
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
CS Colloquium: Dakshita Khurana (University of Illinois, Urbana-Champaign) - Cryptographic Advances in Reasoning about Adversaries
Thu, Mar 02, 2023 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Dakshita Khurana , University of Illinois, Urbana-Champaign
Talk Title: Cryptographic Advances in Reasoning about Adversaries
Series: CS Colloquium
Abstract: A key challenge in cryptography is to ensure that a protocol resists all computationally feasible attacks, even when an adversary decides to follow a completely arbitrary and unpredictable strategy.
This often turns out to be notoriously difficult -- for example, proofs of security must typically extract an adversary's implicit input, but this is at odds with other goals like privacy, which require that inputs be hidden and difficult to extract.
In this talk, I will describe my work that reimagines how we reason about adversaries, thereby settling foundational questions in classical and quantum protocol design. On the classical front, these insights enable efficient verification of computations while preserving privacy, and immunize protocols against coordinated attacks on the internet. On the quantum front, these methods help exploit the "destructive" nature of measurements and open up fundamentally new possibilities for cryptography. I will discuss examples that leverage quantum information to (1) weaken the assumptions needed for core tasks like secure computation on distributed private data, and (2) allow outsourcing computations on sensitive data while also verifying that data was deleted after processing.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Dakshita Khurana is an Assistant Professor of Computer Science at the University of Illinois, Urbana-Champaign. Her research focuses on cryptography and its interactions with quantum information. She has made several contributions to secure protocol design, including to succinct and zero-knowledge proof systems, non-malleable protocols and secure computation. Her work has also impacted fields beyond cryptography, e.g., by establishing the hardness of finding Nash equilibria under standard lattice assumptions. Her recent research enabling secure computation from weak cryptographic structure in the quantum regime was invited as one of the (long) plenary talks at QIP.
Her research has also been recognized via invitations to the SIAM Journal on Computing, awarded to a select few papers at STOC and FOCS.
Dakshita is a recipient of the NSF CAREER award, Visa Research faculty award, and a Graduate of Last Decade (GOLD) Alumni award from IIT-Delhi. In addition, her work has been funded through grants and gifts from the NSF, DARPA, C3AI and Jump Arches. She was named to Forbes List of 30 under 30 in Science and awarded a Google Research Fellowship at the Simons Institute, Berkeley. Her thesis work was previously recognized with a UCLA Dissertation Year Fellowship, a UCLA CS Outstanding PhD Student Award and Outstanding Graduate Awards from Symantec and CISCO.
Host: Jiapeng Zhang
Location: Olin Hall of Engineering (OHE) - 132
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
CS Colloquium: Matus Telgarsky (University of Illinois, Urbana-Champaign) - Searching for the implicit bias of deep learning
Tue, Mar 07, 2023 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Matus Telgarsky, University of Illinois, Urbana-Champaign
Talk Title: Searching for the implicit bias of deep learning
Series: CS Colloquium
Abstract: What makes deep learning special --- why is it effective in so many settings where other models fail? This talk will present recent progress from three perspectives. The first result is approximation-theoretic: deep networks can easily represent phenomena that require exponentially-sized shallow networks, decision trees, and other classical models. Secondly, I will show that their statistical generalization ability --- namely, their ability to perform well on unseen testing data --- is correlated with their prediction margins, a classical notion of confidence. Finally, comprising the majority of the talk, I will discuss the interaction of the preceding two perspectives with optimization: specifically, how standard descent methods are implicitly biased towards models with good generalization. Here I will present two approaches: the strong implicit bias, which studies convergence to specific well-structured objects, and the weak implicit bias, which merely ensures certain good properties eventually hold, but has a more flexible proof technique.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Matus Telgarsky is an assistant professor at the University of Illinois, Urbana-Champaign, specializing in deep learning theory. He was fortunate to receive a PhD at UCSD under Sanjoy Dasgupta. Other highlights include: co-founding, in 2017, the Midwest ML Symposium (MMLS) with Po-Ling Loh; receiving a 2018 NSF CAREER award; and organizing two Simons Institute programs, one on deep learning theory (summer 2019), and one on generalization (fall 2024).
Host: Vatsal Sharan
Location: Olin Hall of Engineering (OHE) - 132
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
CAIS Seminar: Andrew Zolli (Planet) - Using Space and AI to Help Life on Earth: How AI and Satellites Are Transforming Our Stewardship of the Planet
Tue, Mar 07, 2023 @ 04:00 PM - 05:20 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Andrew Zolli, Planet
Talk Title: Using Space and AI to Help Life on Earth: How AI and Satellites Are Transforming Our Stewardship of the Planet
Series: CS Colloquium
Abstract: We're in the middle of two concurrent and convergent technological revolutions. The first is a sensor revolution, in which new streams of real-time data from the ground, the air, and space are making the change on Earth more transparent than ever before. New generations of satellites monitor every crop, every forest, every city, everywhere, every day - and provide unprecedented transparency. The second revolution is an AI summer, in which the wide availability of machine learning, cloud storage and computing are enabling the extraction of real-time indicators from these data sets. This is revealing real-time feedback loops that can show us how our actions impact the world -“ both positively and negatively - and enabling entirely new ways of seeing, analyzing, and responding to planetary change.
In this talk, Planet's Chief Impact Officer Andrew Zolli will share how these breakthrough approaches are transforming our stewardship of the planet, and where they are likely to go next.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: I currently oversee Sustainability and Global Impact initiatives at Planet, a breakthrough space and AI organization that has deployed the largest constellation of Earth-observing satellites in history. These satellites image our whole planet every day in high resolution, and my team makes sure this data is ethically used to its highest and best purposes to accelerate climate action, monitor the world's ecosystems, improve humanitarian action and disaster response, protect human rights, transform sustainable development, advance scientific discovery and artistic expression. We're even exploring how these tools can inform the next iteration of capitalism, where social and environmental externalities are more effectively measured and valued. I also currently serve on the International Board of Directors of Human Rights Watch.
Host: USC Center for Artificial Intelligence in Society (CAIS)
Location: Seeley G. Mudd Building (SGM) - 124
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
CS Colloquium: Alexander Rodríguez (Georgia Tech) - AI for Public Health: Epidemic Forecasting Through a Data-Centric Lens
Thu, Mar 09, 2023 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Alexander RodrÃguez , Georgia Tech
Talk Title: AI for Public Health: Epidemic Forecasting Through a Data-Centric Lens
Series: CS Colloquium
Abstract: Epidemic forecasting is a crucial tool for public health decision making and planning. There is, however, a limited understanding of how epidemics spread, largely due to other complex dynamics, most notably social and pathogen dynamics. With the increasing availability of real-time multimodal data, a new opportunity has emerged for capturing previously unobservable facets of the spatiotemporal dynamics of epidemics. In this regard, my work brings a data-centric perspective to public health via methodological advances in AI at the intersection of time series analysis, spatiotemporal mining, scientific ML, and multi-agent systems. Toward realizing the potential of AI in public health, I addressed multiple challenges stemming from the domain such as data scarcity, distributional changes, and issues arising from real-time deployment to enable our support of CDC's COVID-19 response. This talk will cover methods to address these challenges with novel deep learning architectures for real-time response to disease outbreaks and new techniques for end-to-end learning with mechanistic epidemiological models-”based on differential equations and agent-based models-”that bridge ML advances and traditional domain knowledge to leverage individual merits. I will conclude by discussing challenges and opportunities in public health for data and computer scientists.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Alexander RodrÃguez is a PhD candidate in the College of Computing at Georgia Tech, advised by Prof. B. Aditya Prakash. His research is at the intersection of machine learning, time series, and scientific modeling, and his main application domains are public health and community resilience. He has published at top venues such as AAAI, NeurIPS, ICLR, KDD, WWW, AAMAS, PNAS and has organized workshops and tutorials at AAAI and KDD. His work won the best paper award at ICML AI4ABM 2022 and was awarded the 1st place in the Facebook/CMU COVID-19 Challenge and the 2nd place in the C3.ai COVID-19 Grand Challenge. He was also invited to the Heidelberg Laureate Forum in 2022, and named a 'Rising Star in Data Science' by the University of Chicago Data Science Institute in 2021 and a 'Rising Star in ML & AI' by the University of Southern California in 2022.
Host: Bistra Dilkina
Location: Olin Hall of Engineering (OHE) - 132
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
CS Colloquium: Rika Antonova (Stanford University) - Enabling Self-sufficient Robot Learning
Mon, Mar 20, 2023 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Rika Antonova, Stanford University
Talk Title: Enabling Self-sufficient Robot Learning
Series: CS Colloquium
Abstract: Autonomous exploration and data-efficient learning are important ingredients for helping machine learning handle the complexity and variety of real-world interactions. In this talk, I will describe methods that provide these ingredients and serve as building blocks for enabling self-sufficient robot learning.
First, I will outline a family of methods that facilitate active global exploration. Specifically, they enable ultra data-efficient Bayesian optimization in reality by leveraging experience from simulation to shape the space of decisions. In robotics, these methods enable success with a budget of only 10-20 real robot trials for a range of tasks: bipedal and hexapod walking, task-oriented grasping, and nonprehensile manipulation.
Next, I will describe how to bring simulations closer to reality. This is especially important for scenarios with highly deformable objects, where simulation parameters influence the dynamics in unintuitive ways. The success here hinges on finding a good representation for the state of deformables. I will describe adaptive distribution embeddings that provide an effective way to incorporate noisy state observations into modern Bayesian tools for simulation parameter inference. This novel representation ensures success in estimating posterior distributions over simulation parameters, such as elasticity, friction, and scale, even for scenarios with highly deformable objects and using only a small set of real-world trajectories.
Lastly, I will share a vision of using distribution embeddings to make the space of stochastic policies in reinforcement learning suitable for global optimization. This research direction involves formalizing and learning novel distance metrics on this space and will support principled ways of seeking diverse behaviors. This can unlock truly autonomous learning, where learning agents have incentives to explore, build useful internal representations and discover a variety of effective ways of interacting with the world.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Rika is a postdoctoral scholar at Stanford University and a recipient of the NSF/CRA Computing Innovation Fellowship for research on active learning of transferable priors, kernels, and latent representations for robotics. Rika completed her Ph.D. work on data-efficient simulation-to-reality transfer at KTH. Earlier, she obtained a research Master's degree from the Robotics Institute at Carnegie Mellon University, where she developed Bayesian optimization methods for robotics and for personalized tutoring systems. Before that, Rika was a software engineer at Google, first in the Search Personalization group and then in the Character Recognition team (developing open-source OCR engine Tesseract).
Host: Jesse Thomason
Location: Ronald Tutor Hall of Engineering (RTH) - 115
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
CS Colloquium: Yue Zhao (CMU) - Scalable and Automated Systems and Algorithms for Unsupervised ML
Tue, Mar 21, 2023 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Yue Zhao, Carnegie Mellon University
Talk Title: Scalable and Automated Systems and Algorithms for Unsupervised ML
Series: CS Colloquium
Abstract: Many real-world events do not have outcome labels. For example, the fraudulence of a transaction remains unknown until it is discovered. This is where unsupervised machine learning (ML) becomes crucial in real-world scenarios as it can make decisions based solely on observations. In this talk, I will address two key challenges in unsupervised ML: (i) developing scalable learning systems that can handle large amounts of data, and (ii) automating the selection of the best ML model. The first part of the talk will cover an ML system called TOD, which can "compile" a diverse group of ML algorithms for GPU acceleration. The second part will describe an automated algorithm called MetaOD, which can select top ML models for various applications without relying on labels or evaluations. Lastly, I will discuss my future plans, including the ML+X initiative, which aims to bring the advantages of ML systems and automation to other domains, and the creation of a fully automated ML pipeline that chooses hardware, systems, and models seamlessly.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Yue Zhao is a Ph.D. candidate at CMU, working with Prof. Leman Akoglu and Prof. Zhihao Jia. He focuses on creating scalable and automated ML systems and algorithms, and has published over 30 papers in top venues such as VLDB, MLSys, JMLR, and NeurIPS. His open-source systems (https://github.com/yzhao062) have been widely deployed in firms and industries such as Morgan Stanley and Tesla, and have received over 15,000 GitHub stars and 10 million downloads. Yue has received the CMU Presidential Fellowship and Norton Graduate Fellowship. More information about him can be found at https://www.andrew.cmu.edu/user/yuezhao2/.
Host: Robin Jia
Location: Olin Hall of Engineering (OHE) - 132
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
CS Colloquium: Lindsay Sanneman (MIT) - Transparent Value Alignment: Foundations for Human-Centered Explainable AI in Alignment
Wed, Mar 22, 2023 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Lindsay Sanneman , MIT
Talk Title: Transparent Value Alignment: Foundations for Human-Centered Explainable AI in Alignment
Series: CS Colloquium
Abstract: Alignment of robot objectives with those of humans can greatly enhance robots' ability to act flexibly to safely and reliably meet humans' goals across diverse contexts from space exploration to robotic manufacturing. However, it is often difficult or impossible for humans, both expert and non-expert, to enumerate their objectives comprehensively, accurately, and in forms that are readily usable for robot planning. Value alignment is an open challenge in artificial intelligence that aims to address this problem by enabling robots and autonomous agents to infer human goals and values through interaction. Providing humans with direct and explicit feedback about this value learning process through approaches for explainable AI (XAI) can enable humans to more efficiently and effectively teach robots about their goals. In this talk, I will introduce the Transparent Value Alignment (TVA) paradigm which captures this two-way communication and inference process and will discuss foundations for the design and evaluation of XAI within this paradigm. First, I will present a novel suite of metrics for assessing alignment which have been validated through human subject experiments by applying approaches from cognitive psychology. Next, I will discuss the Situation Awareness Framework for Explainable AI (SAFE-AI), a human factors-based framework for the design and evaluation of XAI across diverse contexts including alignment. Finally, I will propose design guidance for XAI within the TVA context which is grounded in results from a set of human studies comparing a broad range of explanation techniques across multiple domains. I will additionally highlight how this research relates to real-world robotic manufacturing and space exploration settings that I have studied. I will conclude the talk by discussing the future vision of this work.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Lindsay Sanneman is a final year PhD candidate in the Department of Aeronautics and Astronautics at MIT and a member of the Interactive Robotics Group in the Computer Science and Artificial Intelligence Laboratory (CSAIL). Her research focuses on the development of models, metrics, and algorithms for explainable AI (XAI) and AI alignment in complex human-autonomy interaction settings. Since 2018, she has been a member of MIT's Work of the Future task force and has visited over 50 factories worldwide alongside an interdisciplinary team of social scientists and engineers in order to study the adoption of robotics in manufacturing. She is currently a Siegel Research Fellow and has presented her work in diverse venues including the Industry Studies Association and the UN Department of Economic and Social Affairs.
Host: Heather Culbertson
Location: Ronald Tutor Hall of Engineering (RTH) - 109
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
CS Colloquium: Benjamin Eysenbach (CMU) - Self-Supervised Reinforcement Learning
Thu, Mar 23, 2023 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Benjamin Eysenbach , CMU
Talk Title: Self-Supervised Reinforcement Learning
Series: CS Colloquium
Abstract: Reinforcement learning (RL) promises to harness the power of machine learning to solve sequential decision making problems, with the potential to enable applications ranging from robotics to chemistry. However, what makes the RL paradigm broadly applicable is also what makes it challenging: only limited feedback is provided for learning to select good actions. In this talk, I will discuss how we have made headway of this challenge by designing self-supervised RL methods, ones that can learn representations and skills for acting using unsupervised (reward-free) experience. These skill learning methods are practically-appealing and have since sparked a vibrant area of research. I will also share how we have answered some open theoretical questions in this area.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Benjamin Eysenbach is a final-year PhD student at Carnegie Mellon University. His research has developed machine learning algorithms for sequential decision making. His algorithms not only achieve a high degree of performance, but also carry theoretical guarantees, are typically simpler than prior methods, and draw connections between many areas of ML and CS. Ben is the recipient of the NSF and Hertz graduate fellowships. Prior to the PhD, he was a resident at Google Research and studied math as an undergraduate at MIT.
Host: Jyo Deshmukh
Location: Olin Hall of Engineering (OHE) - 132
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
CS Colloquium: Dr. Zhou Li (University of California Irvine) - Debugging the Fragmented DNS Infrastructure at Scale
Thu, Mar 23, 2023 @ 04:00 PM - 05:20 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Dr. Zhou Li, University of California Irvine
Talk Title: Debugging the Fragmented DNS Infrastructure at Scale
Abstract: Domain Name System (DNS) is a fundamental infrastructure that supports almost all sorts of Internet activities. However, service failures and breach of DNS are not rare, and some even led to the shutdown of large data centers, though DNS was designed under the goals like resiliency from the very beginning. We argue that the root causes are that DNS infrastructure has become too fragmented and its protocols have become much more complex, so new research efforts are needed to harden the DNS infrastructure. In this talk, I'll describe our efforts in this direction. First, I'll talk about two new DNS attacks we identified under the settings of domain revocation and conditional resolution, and their implications. Second, I'll talk about how we measure the operational status of DNS-over-Encryption at a large scale. Finally, I'll conclude the talk with an outlook for DNS-related research.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Zhou Li is an Assistant Professor at UC Irvine, EECS department, leading the Data-driven Security and Privacy Lab. Before joining UC Irvine, he worked as Principal Research Scientist at RSA Labs from 2014 to 2018. His research interests include Domain Name System (DNS), Graph Security analytics, Privacy Enhancement Technologies and Side-channel analysis. He received the NSF CAREER award, Amazon Research Award, Microsoft Security AI award and IRTF Applied Networking Research Prize.
Host: Weihang Wang
More Info: https://usc.zoom.us/j/92035174335?pwd=VzhKZ0xjM3A2SzFwOWsyRG1SQWpqUT09
Location: Seeley G. Mudd Building (SGM) - 124
Audiences: Everyone Is Invited
Contact: Melissa Ochoa
Event Link: https://usc.zoom.us/j/92035174335?pwd=VzhKZ0xjM3A2SzFwOWsyRG1SQWpqUT09
-
CS Colloquium: Pavel Izmailov (New York University) - Deconstructing models and methods in deep learning
Mon, Mar 27, 2023 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Pavel Izmailov, New York University
Talk Title: Deconstructing models and methods in deep learning
Series: CS Colloquium
Abstract: Machine learning models are ultimately used to make decisions in the real world, where mistakes can be incredibly costly. We still understand surprisingly little about neural networks and the procedures that we use to train them, and, as a result, our models are brittle, often rely on spurious features, and generalize poorly under minor distribution shifts. Moreover, these models are often unable to faithfully represent uncertainty in their predictions, further limiting their applicability. In this talk, I will present works on neural network loss surfaces, probabilistic deep learning, uncertainty estimation and robustness to distribution shifts. In each of these works, we aim to build foundational understanding of models, training procedures, and their limitations, and then use this understanding to develop practically impactful, interpretable, robust and broadly applicable methods and models.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: I am a final year PhD student in Computer Science at New York University, working with Andrew Gordon Wilson. I am primarily interested in understanding and improving deep neural networks. In particular my interests include out of distribution generalization, probabilistic deep learning, representation learning and large models. I am also excited about generative models, uncertainty estimation, semi-supervised learning, language models and other topics. Recently, our work on Bayesian model selection was recognized with an outstanding paper award at ICML 2022.
Host: Robin Jia
Location: Ronald Tutor Hall of Engineering (RTH) - 115
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
CS Colloquium: Kexin Pei (Columbia University) - Analyzing and Securing Software via Robust and Generalizable Learning
Tue, Mar 28, 2023 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Kexin Pei, Columbia University
Talk Title: Analyzing and Securing Software via Robust and Generalizable Learning
Series: CS Colloquium
Abstract: Software is powering every aspect of our society, but it remains plagued with errors and prone to critical failures and security breaches. Program analysis has been a predominant technique for building trustworthy software. However, traditional approaches rely on hand-curated rules tailored for specific analysis tasks and thus require significant manual effort to tune for different applications. While recent machine learning-based approaches have shown some early promise, they, too, tend to learn spurious features and overfit to specific tasks without understanding the underlying program semantics.
In this talk, I will describe my research on building machine learning (ML) models toward learning program semantics so they can remain robust against transformations in program syntax and generalize to various program analysis tasks and security applications. The corresponding research tools, such as XDA, Trex, StateFormer, and NeuDep, have outperformed commercial tools and prior arts by up to 117x in speed and by 35% in precision and have helped identify security vulnerabilities in real-world firmware that run on billions of devices. To ensure the developed ML models are robust and generalizable, I will briefly describe my research on building testing and verification frameworks for checking the safety properties of deep learning systems. The corresponding research tools, such as DeepXplore, DeepTest, ReluVal, and Neurify, have been adopted and followed up by the industry, been covered in media such as Scientific American, IEEE Spectrum, Newsweek, and TechRadar, and inspired over thousands of follow-up projects.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Kexin Pei is a Ph.D. candidate in Computer Science at Columbia University, advised by Suman Jana and Junfeng Yang. His research lies at the intersection of security, software engineering, and machine learning, with a focus on building machine-learning tools that utilize program structure and behavior to analyze and secure software. His research has received the Best Paper Award in SOSP, an FSE Distinguished Artifact Award, been featured in CACM Research Highlight, and won CSAW Applied Research Competition Runner-Up. He was part of the learning for code team when he interned at Google Brain, building program analysis tools based on large language models.
Host: Jiapeng Zhang
Location: Olin Hall of Engineering (OHE) - 132
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
CS Colloquium: Paul Gölz (Harvard) - Fair, Representative, and Transparent Algorithms for Citizens’ Assemblies
Wed, Mar 29, 2023 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Paul Gölz, Harvard
Talk Title: Fair, Representative, and Transparent Algorithms for Citizens' Assemblies
Series: CS Colloquium
Abstract: Globally, an alternative approach to democracy is gaining momentum: citizens' assemblies, in which randomly selected constituents discuss policy questions and propose solutions. Domain experts have two conflicting requirements on the selection of these assemblies: (1) assemblies should reflect the demographics of the population, and (2) all constituents should have equal chances of being selected. In this talk, I will describe work on designing and analyzing randomized selection algorithms that favorably trade off these objectives. I will share experiences with deploying these algorithms on our online platform Panelot and discuss what we learned from practitioners in the process of adoption. Finally, I will explore how these lessons sparked work on other aspects of citizens' assemblies, such as making the random selection process transparent and managing the discussions within the assembly.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Paul Gölz is a postdoctoral researcher at the School of Engineering and Applied Sciences at Harvard. He received his Ph.D. in computer science from Carnegie Mellon University under the supervision of Ariel Procaccia. Paul studies democratic decision-making and the fair allocation of resources, using tools from algorithms, optimization, and artificial intelligence. Algorithms developed in his work are now deployed to select citizens' assemblies around the world and to allocate refugees for a major US resettlement agency.
Host: David Kempe
Location: Ronald Tutor Hall of Engineering (RTH) - 109
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
CS Colloquium: Emilio Ferrara (USC) - AI & Social Manipulation
Wed, Mar 29, 2023 @ 04:00 PM - 05:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Emilio Ferrara, USC Annenberg / CS
Talk Title: AI & Social Manipulation
Series: CS Colloquium
Abstract: In this talk, I will overview my decadelong journey into understanding the implications of online platform manipulation. I'll start from detecting malicious bots and other forms of manipulation including troll accounts, coordinated campaigns, and disinformation operations. The impact of my work will be corroborated with examples of findings enabled by our technology, e.g., our unveiling of the "Russian bots" operation prior to the 2016 U.S. Presidential election, which informed official Senate investigations and new regulations. I will then illustrate similar issues with the 2020 U.S. Election, as well as COVID-related conspiracies and public health misinformation. I'll conclude by discussing the ML tools we developed to model online mis/disinformation, reveal the malicious adversaries behind the curtains, and characterize their activity, behavior, and strategies, suggesting how they are changing the way researchers and study online platforms in the era of automation and artificial intelligence.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Emilio Ferrara is a professor of communication and computer science at USC Annenberg and at the USC Viterbi Department of Computer Science, professor (by courtesy) of Preventive Medicine at the Keck School, and co-director of the Machine Intelligence and Data Science (MINDS) group at USC ISI. His research focus has been at the intersection between developing theory and methods in network science, machine learning and NLP, and applying them to study socio-technical systems and networks. He is concerned with understanding the implications of AI and networks on human behavior, and their effects on society at large. Ferrara has published 230+ articles that have appeared on venues like the Proceeding of the National Academy of Sciences, Communications of the ACM, Physical Review Letters, and the top ACM, IEEE and AAAI conferences and journals. As a PI at USC, he has received $20M+ in research funding from DARPA, IARPA, NSF, NIH, AFOSR and ONR. Ferrara received accolades including the 2016 DARPA Young Faculty Award and DARPA Director's Fellowship, the 2016 Complex Systems Society Junior Scientific Award, the 2019 USC Viterbi Research Award and the 2022 Research.com Rising Stars award. Until He also served as associate director of the USC Data Science programs.
Host: CS Department
Location: Ronald Tutor Hall of Engineering (RTH) - 526
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
CS Colloquium: Shuang Li (MIT) - Enabling Compositional Generalization of AI Systems
Thu, Mar 30, 2023 @ 11:00 AM - 12:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Shuang Li, Massachusetts Institute of Technology (MIT)
Talk Title: Enabling Compositional Generalization of AI Systems
Series: CS Colloquium
Abstract: A vital aspect of human intelligence is the ability to compose increasingly complex concepts out of simpler ideas, enabling both rapid learning and adaptation of knowledge. Despite their impressive performance, current AI systems fall short in this area and are often unable to solve tasks that fall outside of their training distribution. My research aims to bridge this gap by incorporating compositionality into deep neural networks, thereby enhancing their ability to generalize and solve novel and complex tasks, such as generating 2D images and 3D assets based on complicated specifications, or enabling humanoid agents to perform a diverse range of household activities. The implications of this work are far-reaching, as compositionality has numerous applications across fields such as biology, robotics, and art production. By significantly improving the compositionality ability of AI systems, this research will pave the way for more data-efficient and powerful models in different research areas.
This lecture satisfies requirements for CSCI 591: Research Colloquium
Biography: Shuang Li is a Ph.D. Candidate at MIT, advised by Antonio Torralba. She is interested in developing AI systems that generalize to a wide range of novel tasks and continually learn from the environment. Her research explores methods to incorporate compositionality into deep learning models, giving rise to stronger generalization abilities for solving more challenging novel tasks. Her research involves Generative Modeling, Embodied AI, and Vision-Language Understanding. Shuang is a recipient of the Meta Research Fellowship, Adobe Research Fellowship, MIT Seneff-Zue CS Fellowship, EECS Rising Star, ICML Outstanding Reviewer, and best and outstanding paper awards at NeurIPS workshops.
Host: Swabha Swayamdipta
Location: Olin Hall of Engineering (OHE) - 132
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
