How to write in English (2)
The last decade has witnessed
an experimental revolution in data science and machine learning, epitomised by
deep learning methods.
Remarkably
, the essence
of deep learning is built from two simple algorithmic principles: …
Remarkably & obviously
This text is concerned with exposing
these regularities through unified geometric principles that can be applied throughout a wide spectrum of applications
.
Deep learning systems are no exception
, and since the early days researchers have adapted neural networks to exploit the low-dimensional geometry arising from
physical measurements.
Throughout our exposition
…
… provide principled way to build future architectures yet to be invented
.
Before proceeding,
it is worth noting that
our work concerns representation learning architectures and exploiting the symmetries of data therein
.
That being said,
we believe
that the principles we will focus on are of significant importance in all of these areas.
Rather
, we study
several well-known architectures in-depth
.
Modern deep learning systems typically operate in the so-called
interpolating regime.
Inductive Bias via
Function Regularity.
Modern machine learning operates
(uses) with large, high-quality datasets …
Universal Approximation, however, does not imply an absence of
inductive bias.
we are looking for the most regular functions within
our hypothesis class.
in many applications with even modest
dimension d …
Fully-connected neural networks define function spaces that enable
more flexible notions of regularity, obtained by considering complexity functions c on their weights.
Whenever possible, we will omit
the range C for brevity
.
the domain Ω is usually endowed
with certain geometric structure and symmetries.
Extending these ideas to other domains such as graphs and manifolds and showing how geometric priors emerge from fundamental principles is the main goal of Geometric Deep Learning and the leitmotif
of our text.
Symmetries are ubiquitous
in many machine learning tasks.
The reason is that if both transformations leave the object invariant, then so does the composition of transformations,
and hence the composition is also a symmetry.
Since these objects will be a centerpiece
of the mathematical model of Geometric Deep Learning, they deserve a formal definition and detailed discussion
an equilateral
triangle
infinitesimal
displacements
For instance, the aforementioned
group of rotational and reflection symmetries of a triangle is the same as the group of permutations of a sequence of three elements.
We shall see numerous
instances of group actions in the following sections.
It maps pairs of connected nodes to pairs of connected nodes, and likewise
for pairs of non-connected nodes.
in computer vision, we might assume that planar translations are exact symmetries. However, the real world is noisy and this model falls short in two ways
. Firstly, … Secondly, …
the setting where the domain $\Omega$ is fixed
, and signals $x\in\mathcal{X}(\Omega)$ are undergoing
deformations.
Its utility
in applications depends on introducing an appropriate deformation cost.
Canonical
instances of this are applications dealing with graphs and manifolds.
we define … as the ensemble
of possible input signals
As it turns out
, …
[1] sBronstein, M. M., Bruna, J., Cohen, T., & Veličković, P. (2021). Geometric deep learning: Grids, groups, graphs, geodesics, and gauges. arXiv preprint arXiv:2104.13478.