Two works already established to be
by the famous painter.
A drawing purported to be a Raphael.
WHO’S WHO
IN THE ARTS
Cutting-edge mathematical
forensics at HKUST is bringing
quantitative insights to art
and literature
Two pictures, both thought to be the work
of Raphael. How can mathematics help to
ascertain if they really are by the famous
painter? This intriguing area is among the
domains of Prof Yang Wang, a specialist in
mathematical forensics and stylometry,
namely quantitative analysis of artistic or
literary style.
To ascertain whether a specific painter
or author created a work, he and his team
are using the combined reach of machine
learning and randomization theory to add
to traditional methods of authentication,
such as observational expertise and
connoisseurship. For example, in visual
art analysis, brush stroke measurements,
texture models, fractal models, and color
palette can be comparedwith other works
by the same painter. In literary stylometry,
average length of sentences, synonym
pairs, and frequency of words, among
other features, can be analyzed.
“The computing power for what
we do was available 15 years ago, but
people simply didn’t think about doing
such analysis,” Prof Wang said. “What
has made the difference is the increasing
sophistication of techniques in statistical
analysis, pattern recognition, signal
processing, and machine learning that
have opened the way for almost everything
to be looked at from the perspective of
big data and data analytics.”
Prof Wang first became interested
in stylometry in 2009 when the
Metropolitan Museum of Art in the US
asked him and a collaborator to examine
a corpus of drawings by the great Flemish
artist Pieter Bruegel the Elder (1525-
1569) and famous imitations to see
whether mathematical techniques could
provide insight into identifying forgeries.
Prof Wang’s team later published their
results in the international journal,
IEEE
Transactions of Pattern Recognition
I’m not an artist or historian,
but this interplay of
mathematics being applied to
areas where few thought
it could be applied is
really exciting
PROF YANG WANG
Chair Professor of Mathematics,
Dean of Science
and Machine Intelligence
. Since joining
HKUST in 2014, he has helped his
team members to develop more novel
mathematical techniques in stylometry.
These include a new randomization
technique to analyze challenging open
cases in authorship authentication, where
the authorship in question is not limited
to a small group of “suspects”.
His team’s research has been published
in or submitted to peer-reviewed journals,
such as
Applied and Computational
Harmonic Analysis
,
Adaptive Data Analysis
,
and
Signal Processing
.
In addition, Prof Wang sees great
possibilities for mathematics to extend
from arts and culture into other areas
in social science and humanities in
the future. “We have been looking at
authorship stylometry but such work
could easily move into other areas, such
as computational rhetoric, sentiment
analysis, text mining, image classification,
even government surveillance and
fingerprint analysis. We view this as an
interdisciplinary area with far-reaching
impact down the road.”
The drawing above is part of a private collection.
Was it drawn by Raphael? To conduct stylometric
analysis, HKUST researchers first extracted
quantitative features from the drawing using
wavelet decomposition and scattering transform.
Similar features were extracted from a number of
drawings known to be genuine Raphael works or
forgeries provided by the collector. Features were
then compared. Although such analysis does not
draw definitive conclusions, their results showed that
the drawing’s style was consistent with a Raphael.
The collector also provided some drawings of
unknown provenance that bore a strong resemblance
to those by Raphael. HKUST researchers showed
stylistically they were inconsistent with Raphael.
Is This a Raphael?
