Call for papers:  special session on "Deep and kernel methods: best of two
worlds" at ESANN 2017

European Symposium on Artificial Neural Networks, Computational
Intelligence and Machine Learning (ESANN 2017)

*26-28 April 2017*, Bruges (Belgium)

Multilayer neural networks have experienced a rebirth in the data
analysis field, displaying impressive results, extending even to
classical artificial intelligence domains, such as game playing, computer
vision, natural language and speech processing. The versatility of such
methods have lead deep (semi)-parametric models to get over
well-established learning methods, like kernel machines or classical
statistical techniques. However, their training is a delicate and costly
optimization problem that raises many practical challenges. On the other
hand, kernel methods usually involve solving a tractable convex problem and
are able to handle non-vectorial data directly, leading to a higher
power. Their main drawback is arguably their complexity being
dependent on the number of data points, both at training and model
evaluation times. A natural and emerging field of research is given by
their hybridization, which can done in many fruitful ways. Many ideas
from the deep learning field can be transferred to the kernel
framework and viceversa.

This special session aims at all aspects of deep architectures,
be theoretical or methodological developments, comparative analyses,
or applications. A special emphasis is given to new ideas to bridge
the gap between the fields of deep and kernel learning, as well as
the understanding of their respective weak and strong points.

 The topics of the session include, but are not limited to,

- Applications of deep architectures in data representation and
  analysis, including structured or non-vectorial inputs or outputs

- Natural language and speech processing; structured relationships among
data; scalability/efficiency of deep neural networks and large-scale kernel

- Heterogeneous data and meta-data; applications in neuroscience, computer
vision, (bio)acoustic signals and mechanisms

- Statistical or stability analysis, visualization of learning,
  generalization bounds

- Novel deep(er) architectures/algorithms for data representation and
  learning (using kernels or not)

- Recursive and iterative kernels and their relation to deep neural

- Emulation of multilayer machines by shallow architectures and vice versa

- Randomized (approximate) feature maps to scale-up kernel methods

- Derivation of efficient layer-by-layer algorithms for training
    such networks; reductions in the computational complexity

- Comparisons of deep architectures to shallow architectures


Submitted papers will be reviewed according to the ESANN reviewing process
and evaluated on their scientific value: originality, technical
correctness, and clarity. Tutorial-like contributions are also welcome
provided they add a new perspective on the field.


Paper submission deadline : 19 November 2016
Notification of acceptance : 31 January 2017
ESANN conference : 26-28 April 2017


Lluís A. Belanche, Marta R. Costa-jussà
Universitat Politècnica de Catalunya (Barcelona, Spain)
UNSUBSCRIBE from this page:
Corpora mailing list

Reply via email to