Fall 2024 Joint Seminar

Description: We focus on applications of high-dimensional data analysis to finance, machine learning, genomics, neuroscience and engineering. The seminar will be held across four universities (UC Berkeley, UC Santa Barbara, Florida State University, and UConn), with lectures by students, faculty, and guest speakers. Students are expected to participate in-person in the weekly cross-campus meetings, which will be connected via zoom. The formal presentation will start at 4:10pm ET / 1:10pm PT, and we will continue with informal discussions and exploration of possibilities for future research. Students are encouraged to join collaborative projects emphasizing studies with empirical and simulated data, and to write up and present their own research.

Zoom link: https://ucsb.zoom.us/my/shkolnik

Please check our Discord channel for more projects!

Meeting time: Friday, 4:10pm ET / 1:10pm PT via Zoom.

Locations:
UConn: AUST 344
FSU: LOV 102
UCSB
UC Berkeley: Evans 639. In-person meeting at 2:10pm PT.

Schedule:

• Aug 30: Organizational meeting and introduction. Slides
• Sep 6:
  • Improving quadratic optimization inputs with techniques with concentration of measure (Lisa) Slides
  • Variance concentration and average pairwise correlation in index covariance matrices (Jacob L) Slides
  • Project groups (sign up by Friday 13 September, link)
  • Optimization and long-only constraints (Alex B - if there’s time. If not, first thing next time)
• Sept 13:
  • Empirical properties of index covariance matrices (Jacob L) Slides
  • JSE with long-only constraints (Alex B) Slides
  • Empirical JSE (Zhouli - if there’s time. If not, first thing next time)
• Sept 20:
  • JSE with long-only constraints (Alex B continuing) Slides
• Sept 27:
  • Empirical JSE (Zhouli) Slides
• Oct 4:
  • Duality, factor models and regimes of random matrix theory (Darwin, Harrison and Rahul) Slides
• Oct 11:
  • Recursive methods in long-only minimum variance portfolio weight computations (Nick) Slides
• Oct 18:
  • Empirical JSE (Jake) Slides
• Oct 25:
  • Measuring the misplacement of data from multi-dimensional scaling (Lucy) Slides
  • Weisfeiler and Leman consider sampling (Abdullah) Slides
• Nov 1:
  • JS Corrections of Neural Networks (Enoch) Slides
• Nov 8:
  • Regimes of Random Matrix Theory (Rahul) Slides
• Nov 15:
  • Analyzing factor-based data in different statistical regimes (Rahul) Slides
• Nov 22:
  • Explicit solution of the long-only minimum variance problem (Ololade) Slides
• Dec 6:
  • Empirical Findings on Spectral Properties of Return Covariance Matrices (Darwin and Jingyuan)

Reading list:

• Concentration of measure: background for graphical models, James Stein for Eigenvectors and everything else:
  Foundations of Data Science
  Convex Geometric Analysis
  The weak and strong law of large numbers
  Concentration of measure
  A New Look at Independence
• Machine learning and high dimensional statistics::
  Reconciling modern machine-learning practice and the classical bias–variance trade-off
• James Stein for Eigenvectors / Optimization:
  The Dispersion Bias
  Better Betas
  James–Stein estimation of the first principal component
  James–Stein for the leading eigenvector
  Portfolio optimization via strategy-specific eigenvector shrinkage
  Portfolio Selection Revisited in memory of Harry Markowitz
• Beamforming:
  Robust Minimum Variance Beamforming
• Graphical models:
  Implicit Self-Regularization in Deep Neural Networks: Evidence from Random Matrix Theory and Implications for Learning
  The Application of Improved Covariance Estimation to Adaptive Beamforming and Detection
  On graphical models and convex geometry
  Resolving power and sensitivity to mismatch of optimum array processors
  Multiple Anchor Point Shrinkage for the Sample Covariance Matrix