Latest update

- Scientific Computing on AMD CPUs?
- Why replicating the higher bits of RGB565 when converting to RGBA8888?
- Attempting to connect to VPN
- Broadcom 4360 Wifi (2013 MacBook Air) on Juno?
- Default browser can't play youtube videos in full HD
- Could the President abolish the Supreme Court?
- Could the Senate abolish the Supreme Court?
- My friend needs me to act as a current landlord
- Hagrid, a kind giant
- The Atlas Continuity Issue
- For civiEvents in a Drupal View, how to display only one of a repeating event?
- Issue with CiviMail changing hyperlink urls, causing error at resulting link on iphone but not other devices
- What should I do about this cap iron japanning?
- I am really really trying to find the name of this artist and all i can remember is a few music videos
- I'm not sure how to solve this problem(student question)
- What is the script function to select Solid Body in ANSYS Design Modeler?
- Vacuum Rated Sharpie Or Dry Erase Marker/Board?
- What is the limiting factor for a plant's growth?
- What the federal government can do if a state or town became a dictatorship?
- I need flying cities!

# Determinant of $I + LL^T$

2018-03-15 19:36:45

Let $A$ be symmetric positive definite. Given its eigendecomposition $$A=Q\Lambda Q^T$$ it's easy compute the determinant of $I + A$ as

$$I + A= QQ^T +Q\Lambda Q^T = Q(I+\Lambda)Q^T$$which is similar to $$I+A = \text{diag}(1+\lambda_1 ,\dots,1+\lambda_n)$$

Then the determinant is $$\vert I + A \vert= \prod_k (1+\lambda_k)$$

Now, suppose that instead of the eigendecomposition of $A$, we are given the Cholesky factorization $$A=LL^T$$

Is there still a computationally efficient method for calculating $\vert I + LL^T \vert$?