BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:CCM Seminar: Boris Landa (Yale)
DTSTART;VALUE=DATE-TIME:20210915T140000Z
DTEND;VALUE=DATE-TIME:20210915T151500Z
DTSTAMP;VALUE=DATE-TIME:20210926T162400Z
UID:indico-event-2480@indico.flatironinstitute.org
DESCRIPTION:Presenter: Boris Landa\n\nTitle: Standardizing the Spectra o
f Count Data Matrices by Diagonal Scaling\n\nAbstract: A longstanding que
stion when applying PCA is how to choose the number of principal component
s. Random matrix theory provides useful insights into this question by ass
uming a “signal+noise” model\, where the goal is to estimate the rank
of the underlying signal matrix. If the noise is homoskedastic\, i.e. the
noise variances are identical across all entries\, the spectrum of the noi
se admits the celebrated Marchenko-Pastur (MP) law\, providing a simple me
thod for rank estimation. However\, in many practical situations\, such as
in single-cell RNA sequencing (scRNA-seq)\, the noise is far from being h
omoskedastic. In this talk\, focusing on a Poisson data model\, I will pre
sent a simple procedure termed biwhitening\, which enforces the MP law to
hold by appropriately scaling the rows and columns of the data matrix. Asi
de from the Poisson distribution\, this procedure is extended to families
of distributions with a quadratic variance function. I will demonstrate th
is approach on both simulated and experimental data\, showcasing accurate
rank estimation in simulations and excellent fits to the MP law for real s
cRNA-seq datasets.\n\nIf you would like to attend\, please email crampersa
d@flatironinstitute.org for the Zoom link. \nhttps://indico.flatironinsti
tute.org/event/2480/
LOCATION:Via Zoom + 3rd FL Classroom (Hybrid)
URL:https://indico.flatironinstitute.org/event/2480/
END:VEVENT
END:VCALENDAR