Program
08h45 David Ginsbourger (Opening talk on small data)
09h10 Antoine Guisan Keynote lecture
Species distribution models are statistical tools allowing predictions of species distributions, based on quantification of habitat suitability picturing species' environmental niches and identifying where suitable conditions for species occur in a landscape. I will present these models and review recent advances in this field, and finally show, as novel exciting perspectives, how the use of artificial data through virtual ecologist approaches can help us better understand strengths and weaknesses of these models.
10h00 Spotlight session
1- Michael Barry: "Novelty Search Applied to Small Data in Parameter Tuning"
2- Gülcan Can: "How to Tell Ancient Signs Apart? Recognizing Maya Glyphs with Convolutional Neural Networks"
3- Nam Le: "Improving speech embedding using crossmodal transfer learning with audio-visual data"
4 -Alan Maître: "Challenges in statistical modelling of extreme meteorological events under moderate samples of maxima"
5 -Sebastian Otalora: "Deep Learning in small medical datasets: An active learning approach"
10h30 Coffee break
11h00 Roger Schaer
Image-based precision medicine has tremendous potential for revealing and monitoring disease status in a noninvasive fashion. Recent papers in the field of radiomics showed strong evidence that novel quantitative imaging biomarkers based on structural tissue properties have the potential to complement and even surpass invasive and costly biopsy-based molecular assays in certain clinical contexts. Unfortunately, a majority of the identified imaging biomarkers are perceived as black boxes by end-users, hindering their acceptance in clinical and research environments.
We will present a suite of plugins for the open-access cloud-based ePAD platform enabling the exploration and validation of imaging biomarkers in a clinical environment. The latter include the extraction, visualization and comparison of intensity- and texture-based quantitative imaging features, as well as the construction, use and sharing of user-personalized machine learning models trained with small data. These tools are expected to allow clinicians to quickly identify the meaning and relevance of specific imaging biomarkers in their field of expertise.
11h25 Florian Evequoz (Small Data, Big Picture. Visualizing data to get insights)
11h50 Sylvain Calinon
Human-centered robot applications require the robots to learn new skills by interacting with the end-users. From a machine learning perspective, the challenge is to acquire skills from only few interactions, with strong generalization demands. It requires: 1) the development of intuitive active learning interfaces to acquire meaningful demonstrations; 2) the development of models that can exploit the structure and geometry of the acquired data in an efficient way; 3) the development of adaptive control techniques that can exploit the learned task variations and coordination patterns. The developed models often need to serve several purposes (recognition, prediction, online synthesis), and be compatible with different learning strategies (imitation, emulation, exploration). For the reproduction of skills, these models need to be enriched with force and impedance information to enable human-robot collaboration and to generate safe and natural movements.
I will present an approach combining model predictive control and statistical learning of movement primitives in multiple coordinate systems. The approach will be illustrated in various applications, with robots either close to us (robot for dressing assistance), part of us (prosthetic hand with EMG and tactile sensing), or far from us (teleoperation of bimanual robot in deep water).
12h15 Calixte Mayoraz (Decision Tree vs. Deep learning for image recognition of a small dataset)
12h40 Closing remarks by the organizers followed by a lunch
Document Actions