Connectionists: Special Session at ESANN 2027: Where Kernels Meet Networks

[Frank-Michael Schleif]

*** Apologies for cross postings ***

Dear colleagues,

we would like to draw your attention to the following special session at
ESANN 2027:

*Where Kernels Meet Networks: Neural Tangent Kernels, Gaussian Processes
and Beyond*

*Organizers:*
Frank-Michael Schleif, Technical University of Applied Sciences
Würzburg-Schweinfurt, Germany
Nils M. Kriege, University of Vienna, Austria
Johan Suykens, KU Leuven, Belgium

Kernel methods and neural networks have historically often been viewed as
competing paradigms. A substantial body of recent work, however, has
revealed increasingly close connections between them.

Infinitely wide neural networks can be described in terms of kernels and
Gaussian processes, while Neural Tangent Kernels provide an
infinite-dimensional kernel perspective on the training dynamics of wide
networks. At the same time, current research increasingly moves beyond
fixed limiting kernels towards learnable, structured and expressive kernel
models, finite-width effects, alternative infinite-width limits, and
scalable computational approaches.

The aim of this special session is to bring together these different
perspectives and to encourage interaction between researchers working on
kernel methods, Gaussian processes, neural-network theory, and related
areas.

We welcome theoretical, methodological and applied contributions on topics
including, but not limited to:

   - infinite-width, arc-cosine, compositional and deep kernels;
   - correspondences between neural networks, Gaussian processes and Neural
   Tangent Kernels;
   - finite-width effects and deviations from the NTK regime;
   - feature learning and mean-field limits beyond lazy training;
   - deep kernel processes and learning kernel representations from data;
   - heavy-tailed and stable infinite-width limits;
   - deep and restricted kernel machines;
   - graph kernels, graph neural networks and message-passing expressivity;
   - random features, Nyström methods, low-rank approximations and
   sketching;
   - indefinite, non-metric and structured kernels;
   - Gaussian processes for deep models, uncertainty quantification and
   calibration;
   - applications to molecular, graph, scientific and other structured data.

*Paper submission deadline:* 18 November 2026
*ESANN 2027:* 21–23 April 2027, Bruges, Belgium and online

Submissions to special sessions follow the same review procedure, format
and submission rules as regular ESANN papers. Authors should indicate the
corresponding special session when submitting.

Further information about the special session and submission procedure can
be found on the ESANN website.

We would be very happy to see contributions addressing the increasingly
rich interface between kernel learning, Gaussian processes and neural
networks.

Best regards,

Frank-Michael Schleif
on behalf of the session organizers
Nils M. Kriege · Johan Suykens · Frank-Michael Schleif
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/connectionists/attachments/20260908/a129ec96/attachment.html>