reminder: this is happening at 11am today.
congratulations also to Nicole on her Scientific American article
<https://www.scientificamerican.com/article/quantum-steampunk-19th-century-science-meets-technology-of-today/>
!
On Tue, Apr 21, 2020 at 8:09 PM Aram Harrow <aram(a)mit.edu> wrote:
Dear quanta,
This Friday at 11am we will have Eric tell us about his recent work.
Please everyone also think about any updates you want to mention for the
group, and if you want to be a future speaker or want to nominate someone.
Here are the meeting link and talk details.
https://mit.zoom.us/j/95376219117
Title: Quantum Correlations Can Enhance Generative Modeling
Abstract: In the past few decades, the field of machine learning has grown
substantially and has achieved remarkable success. In addition, quantum
systems can produce complicated probability distributions which are hard to
capture with classical generative models—thus, the field of quantum machine
learning has recently emerged and is expected to facilitate, in particular,
unsupervised learning. However, the origin of the quantum advantage in the
expressivity of quantum machine learning models, although supported by
evidence from computational complexity theory, still remains unclear.
Motivated by this, we consider a particular minimal quantum extension of
Bayesian networks, and prove directly that the presence of quantum
nonlocality and contextuality in this model give rise to a quantum
advantage over the underlying Bayesian network. We also give a numerical
demonstration of a separation between the models on standard machine
learning data sets.
_______________________________________________
qip mailing list
qip(a)mit.edu