Nanospike Workshop – 15-17 May 2019 – IRCICA, Lille (F)
« Spiking Neuron Networks, Nanodevices, Biodiversity »
________________________________________________
Organized by the LAMIA laboratory of the French West Indies University,
funded by the ANR project « NanoSpike », ANR-17-MRS5-0023-01
|
|
|
|
Presentation
Efficient learning with Spiking Neuron Networks (SNN) is the much anticipated counterpart to the currently huge success of Deep Learning which suffer from the drawback of exploding computing ressource consumption. Only SNNs carry the hope to be effectively and efficiently implemented on nanodevices, with a tremendous impact on life sciences and environmental intelligence. SNNs are particularly relevant for temporal pattern recognition, e.g. biosonar signals sent by sea mammals appear like a series of chirps which are eminently suitable to match with the activity patterns of an SNN.
Following the BioComp'2019 meeting, at the same place in Lille, the Nanospike Workshop aims at gathering actors in multiple disciplines and especially sensitizing the community of researchers on bio-inspired devices to the potential of SNNs and their applications.
The Workshop will consist of a challenge, invited talks and poster sessions around the following research topics:
Computing with Spiking Neuron Networks (SNN)
Nanodevices implementing SNN
SNN for Bioacoustics
Call for posters
Any work related to the topics of the Nanospike Workshop, already published or not, might be subject to a poster display and a spot-light oral presentation. The poster submission is open until April 30th and consists of a maximum 1 page abstract (included title, authors and bibliography). Abstracts have to be sent by mail at spike@univ-antilles.fr . Poster submissions will be reviewed and a notification of acceptance will be sent before May 5th.
Call for challenge
From series of echolocation clicks recorded by an hydrophone, the challengers will have to propose a machine learning method and a model for automatic estimation of the range of a sea mammal. Proposals might be received up to May 12th (see Annex for problem definition, link to the data and how to submit).
The challenge participants are strongly encouraged to also submit a poster.
Scheduling
Wednesday May 15th, 14-16 Challenge on Bioacoustics: presentation and demos
Wednesday May 15th, 16-18 Poster session
Thursday May 16th, 9.30-10.30 Acoustics in environmental intelligence and life sciences
Thursday May 16th, 11-12 Learning and computing with Spiking Neuron Networks
Thursday May 16th, 14-15 Spiking neuromorphic computation and systems
Thursday May 16th, 15.30-16.30 Artificial neurons and synapses
Thursday May 16th, 17-18 Nanodevices for implementing Spiking Neuron Networks
Friday May 17th, 9-11:
The Nanospike Workshop is organized in the context of an ANR project, the objective of which is to gather an H2020 European research network. To this end, a private meeting will be held on Friday May 17th, 9-11, with the objective of answering an European call on topics related to the Nanospike Workshop: open to interested people.
Registration
Attendance is free, but every participant has to register before May 12th by sending a mail to spike@univ-antilles.fr .
Deadline recap
Poster submission : April 30th
Poster notification : May 5th
Challenge submission : May 12th
Free registration : May 12th
Location
IRCICA – 50 Avenue Halley – F-59650 Villeneuve-d’Ascq, Lille – France
Hotels: http://gdr-biocomp.fr/wp-content/uploads/2019/03/liste_hotels.pdf
Organization committee
Hélène Paugam-Moisy – LAMIA, Université des Antilles, Guadeloupe – France
Manuel Clergue – LAMIA, Université des Antilles, Guadeloupe – France
Philippe Devienne – IRCICA, Villeneuve-d’Ascq, Lille – France
Pierre Boulet, CRIStAL – Université de Lille – France
Hervé Glotin – LSIS, Université de Toulon – France
Sander Bohte – CWI and Universiteit van Amsterdam, Amsterdam – Netherlands
ANNEX: Challenge
Learn multipulsed clicks to estimate the range of a cetacean
from only one hydrophone
Detecting and localizing the echolocation clicks of cetacean provides insight into its behavior, but usual methods are limited in range, or costly. In summer 2018, DYNI LIS CNRS Toulon University, demonstrated the first 3D tracking of deep diving cetacean from a five-channel small-aperture hydrophone array fixed under a moving autonomous surface vehicle (the ASV Sphyrna, SeaProven SA). The resulting 3D tracks depict the behavior of the cetacean in the abyss (at 1.2 km deep) during 3 hours, with one position at each pulsed emission of the cetacean (Poupard et al ICASSP 2019 – available from the challenge URL). Based on this unique data set, we propose to train and test original mono-channel transient analysis model, that may be adapted to code a sequence of pulses of the biosonar of the whale.
Then the task is to predict with only one channel (sampling rate is 300 kHz, 16 bits) the range of the cetacean for each emitted click. Any kind of machine learning is allowed. The training set has around 4000 clicks. The test set has around 2000 clicks. All the clicks (train and test) are pre-detected and given (raw data) in a matrix (a row = a click). The annotation for each click of the train set = [ recording click Id, x, y, z, and range of the whale ]. All this information can be used during the training.
Your prediction will be formatted in a .csv matrix giving for each click of the test set : [ click Id, estimated range of the whale ].
Your scripts will be sent with your prediction (in Python or Matlab), to check that the model is full automatic. Up to three runs are allowed per challenger. Ranking of the runs will be based on MSE of the ranges.
Deadline : 12th of May, 22:00, Paris time.
Data, details, updates : http://sabiod.lis-lab.fr/workspace/challenges/learn_from_clicks/
Content of the challenge URL:
a link toward the ICASSP paper: Glotin_etal_UTLN_CNRS_IEEEICASSP_2019.pdf
waveform_of_the_clicks_TRAIN_TEST_sets.mat (matlab v7 format, 6036 clicks, 1.5 Go)
README.txt
annotations.txt
Contact & Submission : herve.glotin@univ-tln.fr
Organizers : Hervé Glotin, Maxence Ferrari, Marion Poupard, Pascale Giraudet
