Seminar Listing

Filtering Noise from Correlation/Covariance Matrices. Implications for Trading, Asset Allocation and Risk Management

  • Image
  • October 4, 3:00 - 4:30, Hill Center 552 
  • Abstract: Correlation/Covariance matrices inevitably contain noise due to finite sampling.  It is the typical case that content of the noise component in correlation matrix is more than 90%. Therefore, removal of this noise is of paramount importance since it discriminates “noisy” from “signal” trading, asset allocation, risk management, etc.  A very effective and robust procedure to remove noise from correlation/covariance matrices has been developed.  This procedure exhibits dramatic improvements in (A) various unconstrained and constrained optimization schemes and (B) out-of-sample stability.  The ability of this procedure to detect “regime change” is illustrated on real portfolios.  A few portfolios with obvious underlying correlation structure are presented in order to demonstrate “proof of concept”, i.e. efficacy of the noise filtering procedure. Noise associated with the standard approach to correlation estimation conceals these underlying correlation structures, while the noise filtering procedure exposes them.
  • Speaker: Dr. Alexander Izmailov, Market Memory Trading LLC
  • Bio: Alexander Izmailov has a Ph.D. in Physics (Theoretical and Mathematical Physics, 1986, Lebedev Physics Institute of the former USSR Academy of Sciences). Previously, he had won the Silver Award in the “All Russia” USSR Physics competition among undergraduates. He has attracted $2M in grant money from NASA and NSF (1991-1999, his success rate in grant applications was 1/3). He published more than 40 papers in prestigious scientific journals.  He has held positions as a quantitative trader on vice-president and senior vice-president levels at several major investment banks and hedge funds.  He has developed several successful trading strategies based on innovations in probability theory and statistics.  These strategies were the basis of completely automated trading systems licensed to leading investment banks and hedge funds for trading equities, futures and currency pairs.  He is currently a member of Market Memory Trading, L.L.C.