RSS FeedUpcoming EventsIEOR Seminar Series: Algorithmic Tools for Redistricting — Fairness via Analytics, April 1https://events.berkeley.edu/live/events/243934-ieor-seminar-series-algorithmic-tools-for

Title: Algorithmic Tools for Redistricting: Fairness via Analytics

Abstract: The American winner-take-all congressional district system empowers politicians to engineer electoral outcomes by manipulating district boundaries. To date, computational solutions mostly focus on drawing unbiased maps by ignoring political and demographic input, and instead simply optimize for compactness and other related metrics. However, we maintain that this is a flawed approach because compactness and fairness are orthogonal qualities; to achieve a meaningful notion of fairness, one needs to model political and demographic considerations, using historical data.

We will discuss a series of papers that explore and develop this perspective. In the first (joint with Wes Gurnee), we present a scalable approach to explicitly optimize for arbitrary piecewise-linear definitions of fairness; this employs a stochastic hierarchical decomposition approach to produce an exponential number of distinct district plans that can be optimized via a standard set partitioning integer programming formulation. This enables a large-scale ensemble study of congressional districts, providing insights into the range of possible expected outcomes and the implications of this range on potential definitions of fairness. Further work extending this (joint with Julia Allen & Wes Gurnee), shows that many additional real-world constraints can be easily adapted in this framework (such as minimal county splits as was recently required in Alabama legislation in response to the US Supreme Court decision Milligan v. Alabama). In another paper (joint with Nikhil Garg, Wes Gurnee, and David Rothschild), we study the design of multi-member districts (MMDs) in which each district elects multiple representatives, potentially through a non-winner-takes-all voting rule (as was proposed in H.R. 4000). We carry out large-scale analyses for the U.S. House of Representatives under MMDs with different social choice functions, under algorithmically generated maps optimized for either partisan benefit or proportionality. We find that with three-member districts using Single Transferable Vote, fairness-minded independent commissions can achieve proportional outcomes in every state (up to rounding), and this would significantly curtail the power of advantage-seeking partisans to gerrymander.

/live/events/243934-ieor-seminar-series-algorithmic-tools-for
CLIMB Evergreen talk with Pradeep Ravikumar: The Return of the Kernel Method: Augmentation-Based Self-Supervised Representation Learning & RKHS regression, April 5https://events.berkeley.edu/live/events/239709-climb-evergreen-talk-with-pradeep-ravikumar

The Return of the Kernel Method: Augmentation-Based Self-Supervised Representation Learning & RKHS regression

Abstract: 

Modern self-supervised representation learning approaches such as contrastive learning and masked language modeling have had considerable empirical successes. A typical approach to evaluate such representations involves learning a linear probe on top of such representations and measuring prediction performance with respect to some downstream prediction task. We add to a burgeoning understanding of such representation learning approaches as learning eigenfunctions of certain Laplacian operators, so that learning a linear probe can be naturally connected to RKHS regression with implicitly specified kernels. This allows us to extend non-parametric tools from RKHS regression analysis to analyze the performance of self-supervised representation learning methods, in a way that completely side-steps grappling with neural network based function classes used in practice for the representation encoders. This contravenes prevailing wisdom that we cannot understand these modern representation learning methods without understanding the inductive bias implicit in the intersection of deep neural networks and the optimization methods used to learn them. We specifically focus on augmentation based self-supervised learning approaches, and isolate key structural complexity characterizations of augmentations that we show can be used to quantitatively inform downstream performance of the learned representations.

Bio: 

Pradeep Ravikumar is a Professor in the Machine Learning Department, School of Computer Science at Carnegie Mellon University. His thesis has received honorable mentions in the ACM SIGKDD Dissertation award and the CMU School of Computer Science Distinguished Dissertation award. He is a Sloan Fellow, a Siebel Scholar, a recipient of the NSF CAREER Award, and co-editor-chief of the Journal of Machine Learning Research. His recent research interests are in neuro-symbolic AI, combining statistical machine learning, and symbolic and causal learning.
/live/events/239709-climb-evergreen-talk-with-pradeep-ravikumar
Polar Topological Defects – Fundamentals to Applications: Nano Seminar series, April 5https://events.berkeley.edu/live/events/230393-polar-topological-defects-fundamentals-to

Topological defects such as vortices and skyrmions have recently gained significant interest in solid state materials as ferroic materials (ferromagnets and ferroelectrics) have become a test-bed to realize and control these nanoscale structures. Although this phenomenon is being investigated as a pathway to energy efficient information storage, broader applications in interaction of electromagnetic waves with such features are emerging.

In the case of ferroelectrics, boundary condition engineering is used to achieve vortices, skyrmions, and merons in low dimensional epitaxial oxide heterostructures.

In this talk, I will introduce the notion that similar phenomenology but at the atomic scale can be achieved in charge density wave phases, especially nominally semiconducting chalcogenides. I will outline my group and other groups’ efforts in showing non-trivial toroidal polar topologies at the atomic level in chalcogenides with nominally empty conduction band with d-orbital character such as 1T-TiSe2, Ta2NiSe5 and BaTiS3.

Specifically, we use X-ray single crystal diffraction as a probe for high quality single crystals of a quasi-1D hexagonal chalcogenide, BaTiS3, to reveal complex polar topologies such as vortices, and head-to-head and tail-to-tail arrangement of dipoles. Recent experiments and theoretical studies on the stability and dynamics of these features, and their broad connection to low dimensional magnets, will also be discussed. Lastly, I will outline the perspective for photonic applications of polarization textures.

***********

Jayakanth Ravichandran did his PhD in the Ramesh lab here at UC Berkeley (Go Bears!) and postdoc with Philip Kim at Harvard. He joined the USC faculty in 2015.

/live/events/230393-polar-topological-defects-fundamentals-to
BIDS Seminar with Arash Ardakani, April 9https://events.berkeley.edu/live/events/240909-bids-seminar-with-arash-ardakani

Arash Ardakani gives a 30-minute presentation about his most recent research and leads a 30-minute Q&A.


DiffSampler: A Differential and Inherently Parallel Sampling Method for Verification

“Diverse input samples to software and hardware designs are essential for their thorough testing and verification to ensure reliability, validity, and applicability of the results to real-world scenarios. Generating such samples is a hard computational problem due to the inherent complexity, size of the search space, and resource constraints involved in the process. Addressing these challenges has prompted the development of specialized algorithms that heavily rely on heuristics. Different from such heuristic algorithms, I recently proposed a novel differentiable sampling method, called DiffSampler, that employs gradient descent (GD) to learn diverse input samples. In this talk, I introduce DiffSampler and show how it formulates the verification problem into a supervised multi-output regression task where its loss function is minimized using GD. Such a differentiable method enables performing the learning operations in parallel, leading to GPU-accelerated sampling and accordingly significant throughput improvements over state-of-the-art heuristic samplers.” - Arash Ardakani

/live/events/240909-bids-seminar-with-arash-ardakani
CLIMB Evergreen Talk with Martin Wainwright- Challenges with Covariate Shift: From Prediction to Causal Inference, April 12https://events.berkeley.edu/live/events/243186-climb-evergreen-talk-with-martin-wainwright

Challenges with Covariate Shift: From Prediction to Causal Inference

Abstract:

In many modern uses of predictive methods, there can be shifts between the distributional properties of training data compared to the test data. Such mismatches can cause dramatic reductions in accuracy that remain mysterious. How to find practical procedures that mitigate such effects in an optimal way? In this talk, we discuss the fundamental limits of problems with covariate shift, and simple procedures that achieve these fundamental limits. Our talk covers both the challenges of covariate shift in non-parametric regression,
and also for semi-parametric problems that arise from causal inference and off-policy evaluation.

Based on joint works with: Peter Bartlett, Peng Ding, Cong Ma, Wenlong Mou, Reese Pathak and Lin Xiao.

Bio:

Martin Wainwright is the Cecil H. Green Professor in Electrical Engineering and Computer Science and Mathematics at MIT, and affiliated with the Laboratory for Information and Decision Systems and Statistics and Data Science Center. He joined the MIT faculty in July 2022 after spending 20 years in Statistics and EECS at the University of California at Berkeley.

Martin is broadly interested in statistics, machine learning, information theory and optimization. His work has been recognized by various awards, among them the COPSS Presidents’ Award from the Joint Statistical Societies, the David Blackwell Award from the Institute of Mathematical Statistics, a Section Lecturer from the International Congress of Mathematicians, and a Sloan Foundation Fellowship. He has co-authored several books, including on graphical models, sparse statistical models, and high-dimensional statistics.

/live/events/243186-climb-evergreen-talk-with-martin-wainwright
Historical Data, Present-Day Harms: On the Uses and Limits of Data Science for the Study of Social Movements, April 22https://events.berkeley.edu/live/events/223324-historical-data-present-day-harms-on-the-uses-and

How can data-scientific methods be used to surface the otherwise invisible forms of labor, agency, and action that are embedded in the historical record? How might these methods be adapted to the study of present-day social change? Placing a computational analysis of the nineteenth-century abolitionist movement in dialogue with new work on the language and structure of online social movements, this talk will consider the uses and limits of data-scientific methods when applied to living data and in light of real-world harms.

/live/events/223324-historical-data-present-day-harms-on-the-uses-and
BIDS Seminar with Cody Markelz, PhD, April 23https://events.berkeley.edu/live/events/208851-bids-seminar-with-cody-markelz-phd

Cody Markelz PhD gives a 30-minute presentation about his most recent research and leads a 30-minute Q&A.

Join us in person: BIDS provides lunch!


Data Landscapes: Visual Storytelling of California’s Fiery and Frosty Extremes

Cody’s talk will blend research findings, illustrations, data visualizations, and field-based journalism to explore the impact of fire and avalanches on California’s ecosystems.

/live/events/208851-bids-seminar-with-cody-markelz-phd
BIDS Seminar with Stéfan van der Walt, PhD, April 25https://events.berkeley.edu/live/events/240871-bids-seminar-with-stfan-van-der-walt-phd

Stéfan van der Walt gives a 30-minute presentation about his most recent research and leads a 30-minute Q&A.


Scientific Python: Community, Tools, and Open Science

The scientific Python ecosystem comprises foundational libraries like NumPy and SciPy, technique-specific libraries like NetworkX and scikit-image, and domain-specific libraries such as PyHEP and AstroPy. We present the Scientific Python project, an effort to better coordinate and support the community of scientific Python ecosystem developers. We place the ecosystem within its historic context, explore the radical open transformation witnessed in, and brought about by, data science, and discuss challenges faced in its growth and maintenance.

Bio:

Stéfan van der Walt is a senior research data scientist at BIDS.
He has been developing scientific open-source software for more than fifteen years, focusing primarily on tools in the Python language.
He is the founder of scikit-image, co-author of “Elegant SciPy: The Art of Scientific Python”,
co-architect of SkyPortal (an astronomy data platform for the ZTF2 survey at Caltech), and co-developer of the viridis colormap.

/live/events/240871-bids-seminar-with-stfan-van-der-walt-phd
BIDS Seminar with Antonia Winkler, April 30https://events.berkeley.edu/live/events/240840-bids-seminar-with-antonia-winkler

Invited guest Antonia Winkler gives a 30-minute presentation and leads a 30-minute Q&A. 

Join us in person: BIDS provides lunch!


Open science at CERN: Infrastructure, policy and practice

The high-energy physics research conducted at the European Organization for Nuclear Research (CERN) involves the generation and application of a highly diverse set of scientific resources. CERN’s Large Hadron Collider (LHC), the largest experimental setup in the world, requires the development of complex software and hardware components that enable the generation, storage and analysis of research data on an exabyte scale. Large-scale experimentation at CERN further necessitates the coordination of several thousand researchers and has led to the creation of a unique set of software tools that support the organization of research practices throughout the institution.

The highly complex nature of research at CERN has presented the laboratory with challenges in opening up its diverse research outputs. In the course of CERN’s existence, various efforts have emerged to make software, hardware and data generally accessible. This talk will provide an overview of open science milestones at CERN and point to the institutional structures that support the opening up of the lab’s scientific resources. With its unique strategy of taking both software and hardware into account, the newly established Open Source Program Office (OSPO) will receive special attention in this context. To provide an impression of CERN’s diverse landscape of practitioners, open data initiatives such as the CERN Open Data Portal, open hardware efforts such as White Rabbit, and source software endeavors such as Zenodo will be introduced. Finally, the talk will explore the correlations between open science policy, infrastructure and practice at CERN, with a specific focus on the dual role of open source projects as open science infrastructures and open research outputs in their own right.

/live/events/240840-bids-seminar-with-antonia-winkler
Spring 2024 Campuswide Ceremony, May 11https://events.berkeley.edu/live/events/243373-spring-2024-campuswide-ceremony

Congratulations, Class of 2024! The campuswide Commencement for all undergraduate and graduate students, in every school and college, is on Saturday, May 11, 2024, at 10:30 a.m. at California Memorial Stadium. Check commencement.berkeley.edu for updates on how to have a smooth, memorable experience.

/live/events/243373-spring-2024-campuswide-ceremony
BIDS Seminar with Karthik Ram, PhD, May 21https://events.berkeley.edu/live/events/208996-bids-seminar-with-karthik-ram

Karthik Ram PhD gives a 30-minute presentation about his most recent research and leads a 30-minute Q&A.

Join us in person: BIDS provides lunch!

/live/events/208996-bids-seminar-with-karthik-ram
Inaugural Berkeley-Stanford Workshop on Veridical Data Science, May 31https://events.berkeley.edu/live/events/239397-inaugural-berkeley-stanford-workshop-on-veridical

The Berkeley-Stanford Veridical Data Science Workshop is focused on showcasing and promoting veridical (truthful) data science (VDS) for reproducible, reliable data analysis and decision-making. It intends to build a community of veridical data science researchers for trustworthy data science, machine learning, and artificial intelligence. The discussions will promote opportunities for statisticians and data scientists to identify important VDS research topics and critical applications in academia and industry. Graduate students and early career researchers will benefit from this conference to find future research directions.

Button: Learn More, Buy A Ticket


Organizing Committee:
Bin Yu (UC Berkeley, co-chair), Russ Poldrack (Stanford University, co-chair), Maya Mathur (Stanford University), and Tiffany Tang (University of Michigan)

/live/events/239397-inaugural-berkeley-stanford-workshop-on-veridical