Enhanced Rotational Invariant Convolutional Neural Network for Supernovae Detection

In this paper, we propose an enhanced CNN model for detecting supernovae (SNe). This is done by applying a new method for obtaining rotational invariance that exploits cyclic symmetry. In addition, we use a visualization approach, the layerwise relevance propagation (LRP) method, which allows finding the relevant pixels in each image that contribute to discriminate between SN candidates and artifacts. We introduce a measure to assess quantitatively the effect of the rotational invariant methods on the LRP relevance heatmaps. This allows comparing the proposed method, CAP, with the original Deep-HiTS model. The results show that the enhanced method presents an augmented capacity for achieving rotational invariance with respect to the original model. An ensemble of CAP models obtained the best results so far on the HiTS dataset, reaching an average accuracy of 99.53%. The improvement over Deep-HiTS is significant both statistically and in practice.

[1]  Eric Jones,et al.  SciPy: Open Source Scientific Tools for Python , 2001 .

[2]  Pavlos Protopapas,et al.  Computational Intelligence Challenges and Applications on Large-Scale Astronomical Time Series Databases , 2014, IEEE Computational Intelligence Magazine.

[3]  Wojciech Samek,et al.  Methods for interpreting and understanding deep neural networks , 2017, Digit. Signal Process..

[4]  Alexander Binder,et al.  Evaluating the Visualization of What a Deep Neural Network Has Learned , 2015, IEEE Transactions on Neural Networks and Learning Systems.

[5]  E. Bertin,et al.  SExtractor: Software for source extraction , 1996 .

[6]  Yuan Yu,et al.  TensorFlow: A system for large-scale machine learning , 2016, OSDI.

[7]  Andrew Zisserman,et al.  Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps , 2013, ICLR.

[8]  Pablo A. Estévez,et al.  Supernovae detection by using convolutional neural networks , 2016, 2016 International Joint Conference on Neural Networks (IJCNN).

[9]  F. Förster,et al.  THE HIGH CADENCE TRANSIENT SURVEY (HITS). I. SURVEY DESIGN AND SUPERNOVA SHOCK BREAKOUT CONSTRAINTS , 2016, 1609.03567.

[10]  Sander Dieleman,et al.  Rotation-invariant convolutional neural networks for galaxy morphology prediction , 2015, ArXiv.

[11]  Koray Kavukcuoglu,et al.  Exploiting Cyclic Symmetry in Convolutional Neural Networks , 2016, ICML.

[12]  Alexander Binder,et al.  On Pixel-Wise Explanations for Non-Linear Classifier Decisions by Layer-Wise Relevance Propagation , 2015, PloS one.

[13]  Eduardo Serrano,et al.  LSST: From Science Drivers to Reference Design and Anticipated Data Products , 2008, The Astrophysical Journal.

[14]  Pablo A. Estévez,et al.  Deep-HiTS: Rotation Invariant Convolutional Neural Network for Transient Detection , 2017, ArXiv.

[15]  Jian Sun,et al.  Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification , 2015, 2015 IEEE International Conference on Computer Vision (ICCV).

[16]  Max Welling,et al.  Group Equivariant Convolutional Networks , 2016, ICML.

[17]  Rob Fergus,et al.  Visualizing and Understanding Convolutional Networks , 2013, ECCV.

[18]  Kenneth Patton,et al.  Status of the Dark Energy Survey Camera (DECam) project , 2010, Other Conferences.