Supervised Neural Network Structure Recovery

Ildefons Magrans de Abril, Ann Nowé
; Proceedings of the Neural Connectomics Workshop at ECML 2014, PMLR 46:37-44, 2015.

Abstract

This paper presents our solution to the European Conference of Machine Learning Neural Connectomics Discovery Challenge. The challenge goal was to improve the performance of existing methods for recovering the neural network structure given the time series of neural activities. We propose to approximate a function able to combine several connectivity indicators between neuron pairs where each indicator is the result of running a feature engineering pipeline optimized for a particular noise level and firing synchronization rate among neurons. We proved the suitability of our solution by improving the state of the art prediction performance more than 6% and by obtaining the third best score on the test dataset out of 144 teams.

Cite this Paper


BibTeX
@InProceedings{pmlr-v46-abril15, title = {Supervised Neural Network Structure Recovery}, author = {Ildefons Magrans de Abril and Ann Nowé}, booktitle = {Proceedings of the Neural Connectomics Workshop at ECML 2014}, pages = {37--44}, year = {2015}, editor = {Demian Battaglia and Isabelle Guyon and Vincent Lemaire and Jordi Soriano}, volume = {46}, series = {Proceedings of Machine Learning Research}, month = {15 Sep}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v46/abril15.pdf}, url = {http://proceedings.mlr.press/v46/abril15.html}, abstract = {This paper presents our solution to the European Conference of Machine Learning Neural Connectomics Discovery Challenge. The challenge goal was to improve the performance of existing methods for recovering the neural network structure given the time series of neural activities. We propose to approximate a function able to combine several connectivity indicators between neuron pairs where each indicator is the result of running a feature engineering pipeline optimized for a particular noise level and firing synchronization rate among neurons. We proved the suitability of our solution by improving the state of the art prediction performance more than 6% and by obtaining the third best score on the test dataset out of 144 teams.} }
Endnote
%0 Conference Paper %T Supervised Neural Network Structure Recovery %A Ildefons Magrans de Abril %A Ann Nowé %B Proceedings of the Neural Connectomics Workshop at ECML 2014 %C Proceedings of Machine Learning Research %D 2015 %E Demian Battaglia %E Isabelle Guyon %E Vincent Lemaire %E Jordi Soriano %F pmlr-v46-abril15 %I PMLR %J Proceedings of Machine Learning Research %P 37--44 %U http://proceedings.mlr.press %V 46 %W PMLR %X This paper presents our solution to the European Conference of Machine Learning Neural Connectomics Discovery Challenge. The challenge goal was to improve the performance of existing methods for recovering the neural network structure given the time series of neural activities. We propose to approximate a function able to combine several connectivity indicators between neuron pairs where each indicator is the result of running a feature engineering pipeline optimized for a particular noise level and firing synchronization rate among neurons. We proved the suitability of our solution by improving the state of the art prediction performance more than 6% and by obtaining the third best score on the test dataset out of 144 teams.
RIS
TY - CPAPER TI - Supervised Neural Network Structure Recovery AU - Ildefons Magrans de Abril AU - Ann Nowé BT - Proceedings of the Neural Connectomics Workshop at ECML 2014 PY - 2015/10/21 DA - 2015/10/21 ED - Demian Battaglia ED - Isabelle Guyon ED - Vincent Lemaire ED - Jordi Soriano ID - pmlr-v46-abril15 PB - PMLR SP - 37 DP - PMLR EP - 44 L1 - http://proceedings.mlr.press/v46/abril15.pdf UR - http://proceedings.mlr.press/v46/abril15.html AB - This paper presents our solution to the European Conference of Machine Learning Neural Connectomics Discovery Challenge. The challenge goal was to improve the performance of existing methods for recovering the neural network structure given the time series of neural activities. We propose to approximate a function able to combine several connectivity indicators between neuron pairs where each indicator is the result of running a feature engineering pipeline optimized for a particular noise level and firing synchronization rate among neurons. We proved the suitability of our solution by improving the state of the art prediction performance more than 6% and by obtaining the third best score on the test dataset out of 144 teams. ER -
APA
Abril, I.M.d. & Nowé, A.. (2015). Supervised Neural Network Structure Recovery. Proceedings of the Neural Connectomics Workshop at ECML 2014, in PMLR 46:37-44

Related Material