New Phd grant opens in oct. nov. 2015 (3 years)

New Postdoc grant opend in oct. dec. 2015 (1 year)


in 'Deep Learning for human and non human Bioacoustics'


The goal of your research will be to improve the current performance of soundscape / bioacoustic pattern detection and classification, at low signal to noise ratio, and/or into large data set (Big Data paradigm).

Thus, the objectives here are three-fold: (a) to make the signal representation more robust, (b) to develop classification model more efficient on complex bioacoustic patterns, with supervised but also unsupervised approaches, and (c) to manage and collect additional training data to better model the variability of object categories (within some terrestrial or submarine recording experiments).

The frequent representations are either based on Fourier descriptors or MFCC. However we design mid-level or high-level features based on time-frequency segmentation, wavelet and discrete decomposition, compress sensing, non parametric bayesian representation, CNN / DNN.

The validation of the models are conducted from cetaceans to birds songs, bats to dolphins biosonars, from frogs to insects calls. Biodiversity analysis and environmental care projects are some of the direct outcomes of this research. You'll develop high level research, and you may collaborate with some industry, which could afterward offer you a permanent position in addition to academic opportunities.


Your profile:

  • Machine learning / Data Science or Applied Mathematics / Computer Science, or Electrical Engineering / Signal Processing,
  • Solid mathematics knowledge (especially linear algebra and statistics),
  • Creative and highly motivated, good programming skill (Python, Matlab...),
  • Fluent in English, both written and spoken,
  • Prior knowledge in the areas of pattern recognition and / or neurobiology is a plus.

Location:

Current positions are available on the French Riviera, university of Toulon - CNRS lab LSIS heading SABIOD, France. Toulon campus provides scientific facilities and is located in the beautiful Provence Côte d'Azur area in France, between Marseille and Nice. LSIS lab actually gathers 200 researchers professors and PhD students in computer science and signal analysis and pattern recognition.
Your research will be conducted within collaborations, like National Parks (Port-Cros), international labs (in Italy, USA, Germany, Spain, Canada, Australia, UK...), and may include outdoors signal experiments on cetaceans, birds... Worldwide collaborations offers real opportunities to travel (Reunion, New Caledonia, USA, Canada).


Details


DYNI / academic LSIS lab is seeking a PhD student and a Post doc in Toulon, France, to investigate deep learning for efficient bioacoustic representations, applied to human speech, bird and whales sound analyses and classification. Evaluation will be conducted on international benchmarks. Since 2012, DYNI heads the SABIOD consortium ( http://sabiod.org ) on Scaled Acoustic Biodiversity, joining several international teams expert in machine learning, signal processing and bioacoustics. DYNI chaired several workshops on machine learning meetings bioacoustics at ICML 2013 and 2014, NIPS 2013, and ICDM 2015. DYNI has the chair of Scene Analysis at Inst. Univ. de France (iuf.amue.fr), and th group is dedicated to multimodal scene analysis from signal to automatic indexing. DYNI maintains close ties with theoretical research groups in Paris (DATA team; Sorbonne Univ...), INRIA, Cornell Univ, NYU, Victoria Univ, Pavia Univ., and some deep learning community. DYNI has close ties with experimental research groups in Scaled Bioacoustic: Ocean Network Canada (UVIC), Antares deep sea platform... some ONGs (OrcaLab, XenoCanto), and industrial partnairs.
Applicants should have a strong background in machine learning (maths, statistics, computer science) and a genuine interest in understanding neural computation in perceptual intelligence. Prior exposure to neural computation and machine learning and programming skills is advantageous. Applicants with interest in combining machine learning, feature learning and signal processing are encouraged to apply. Screening of applications will begin sept. 15th. The starting date of the 100% government grant (MRT) is October or at last December. The full Phd grant lasts at least 3 years (french Phd program lasts usually 3 years). The Post doc grant lasts 1 year or more.
Application materials should include a CV (university grades, honors / awards, brief statement of research interests, contact details of 2 referees), + 1 or 2 work samples, anything that is genuinely the own work of the applicant (e.g. thesis, computer code, web site demo, research manuscript / essay).
For informal inquiries, please contact the direction team of this project : Dr. Joseph Razik, Pr. Hervé Glotin and Dr. Sébastien Paris : razik@univ-tln.fr, h.glotin@gmail.com, sebastien.paris@lsis.org

Submit your application (a single .PDF) before the 12th of Sept to:


phd.application.dyni@gmail.com

(ps: Some other positions are available as: Master Intership, Engineers ; spontaneous applications are welcome).