Commit a3e3b424 authored by Samuel Flender's avatar Samuel Flender
Browse files

first commit

parents
File added
This diff is collapsed.
ac3_gt.png

35.4 KB

ac4_gt.png

37.6 KB

@article{DBLP:journals/corr/GoyalDGNWKTJH17,
author = {Priya Goyal and
Piotr Doll{\'{a}}r and
Ross B. Girshick and
Pieter Noordhuis and
Lukasz Wesolowski and
Aapo Kyrola and
Andrew Tulloch and
Yangqing Jia and
Kaiming He},
title = {Accurate, Large Minibatch {SGD:} Training ImageNet in 1 Hour},
journal = {CoRR},
volume = {abs/1706.02677},
year = {2017},
url = {http://arxiv.org/abs/1706.02677},
archivePrefix = {arXiv},
eprint = {1706.02677},
timestamp = {Mon, 03 Jul 2017 13:29:02 +0200},
biburl = {http://dblp.org/rec/bib/journals/corr/GoyalDGNWKTJH17},
bibsource = {dblp computer science bibliography, http://dblp.org}
}
@article{DBLP:journals/corr/abs-1708-02983,
author = {Yang You and
Aydin Bulu{\c{c}} and
James Demmel},
title = {Scaling Deep Learning on {GPU} and Knights Landing clusters},
journal = {CoRR},
volume = {abs/1708.02983},
year = {2017},
url = {http://arxiv.org/abs/1708.02983},
archivePrefix = {arXiv},
eprint = {1708.02983},
timestamp = {Tue, 05 Sep 2017 10:03:46 +0200},
biburl = {http://dblp.org/rec/bib/journals/corr/abs-1708-02983},
bibsource = {dblp computer science bibliography, http://dblp.org}
}
@article{DBLP:journals/corr/abs-1709-05011,
author = {Yang You and
Zhao Zhang and
Cho{-}Jui Hsieh and
James Demmel},
title = {100-epoch ImageNet Training with AlexNet in 24 Minutes},
journal = {CoRR},
volume = {abs/1709.05011},
year = {2017},
url = {http://arxiv.org/abs/1709.05011},
archivePrefix = {arXiv},
eprint = {1709.05011},
timestamp = {Thu, 05 Oct 2017 09:42:47 +0200},
biburl = {http://dblp.org/rec/bib/journals/corr/abs-1709-05011},
bibsource = {dblp computer science bibliography, http://dblp.org}
}
@article{DBLP:journals/corr/JanuszewskiMLKD16,
author = {Michal Januszewski and
Jeremy Maitin{-}Shepard and
Peter Li and
J{\"{o}}rgen Kornfeld and
Winfried Denk and
Viren Jain},
title = {Flood-Filling Networks},
journal = {CoRR},
volume = {abs/1611.00421},
year = {2016},
url = {http://arxiv.org/abs/1611.00421},
archivePrefix = {arXiv},
eprint = {1611.00421},
timestamp = {Wed, 07 Jun 2017 14:42:04 +0200},
biburl = {http://dblp.org/rec/bib/journals/corr/JanuszewskiMLKD16},
bibsource = {dblp computer science bibliography, http://dblp.org}
}
@article {Januszewski200675,
author = {Januszewski, Micha{\l} and Kornfeld, J{\"o}rgen and Li, Peter H and Pope, Art and Blakely, Tim and Lindsey, Larry and Maitin-Shepard, Jeremy B and Tyka, Mike and Denk, Winfried and Jain, Viren},
title = {High-Precision Automated Reconstruction of Neurons with Flood-filling Networks},
year = {2017},
doi = {10.1101/200675},
publisher = {Cold Spring Harbor Laboratory},
abstract = {Reconstruction of neural circuits from volume electron microscopy data requires the tracing of complete cells including all their neurites. Automated approaches have been developed to perform the tracing, but without costly human proofreading their error rates are too high to obtain reliable circuit diagrams. We present a method for automated segmentation that, like the majority of previous efforts, employs convolutional neural networks, but contains in addition a recurrent pathway that allows the iterative optimization and extension of the reconstructed shape of individual neural processes. We used this technique, which we call flood-filling networks, to trace neurons in a data set obtained by serial block-face electron microscopy from a male zebra finch brain. Our method achieved a mean error-free neurite path length of 1.1 mm, an order of magnitude better than previously published approaches applied to the same dataset. Only 4 mergers were observed in a neurite test set of 97 mm path length.},
URL = {https://www.biorxiv.org/content/early/2017/10/09/200675},
eprint = {https://www.biorxiv.org/content/early/2017/10/09/200675.full.pdf},
journal = {bioRxiv}
}
@article{DBLP:journals/corr/RonnebergerFB15,
author = {Olaf Ronneberger and
Philipp Fischer and
Thomas Brox},
title = {U-Net: Convolutional Networks for Biomedical Image Segmentation},
journal = {CoRR},
volume = {abs/1505.04597},
year = {2015},
url = {http://arxiv.org/abs/1505.04597},
archivePrefix = {arXiv},
eprint = {1505.04597},
timestamp = {Wed, 07 Jun 2017 14:40:33 +0200},
biburl = {http://dblp.org/rec/bib/journals/corr/RonnebergerFB15},
bibsource = {dblp computer science bibliography, http://dblp.org}
}
@article{DBLP:journals/corr/CicekALBR16,
author = {{\"{O}}zg{\"{u}}n {\c{C}}i{\c{c}}ek and
Ahmed Abdulkadir and
Soeren S. Lienkamp and
Thomas Brox and
Olaf Ronneberger},
title = {3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation},
journal = {CoRR},
volume = {abs/1606.06650},
year = {2016},
url = {http://arxiv.org/abs/1606.06650},
archivePrefix = {arXiv},
eprint = {1606.06650},
timestamp = {Wed, 07 Jun 2017 14:41:35 +0200},
biburl = {http://dblp.org/rec/bib/journals/corr/CicekALBR16},
bibsource = {dblp computer science bibliography, http://dblp.org}
}
@article{KASTHURI2015648,
title = "Saturated Reconstruction of a Volume of Neocortex",
journal = "Cell",
volume = "162",
number = "3",
pages = "648 - 661",
year = "2015",
issn = "0092-8674",
doi = "https://doi.org/10.1016/j.cell.2015.06.054",
url = "http://www.sciencedirect.com/science/article/pii/S0092867415008247",
author = "Narayanan Kasthuri and Kenneth Jeffrey Hayworth and Daniel Raimund Berger and Richard Lee Schalek and José Angel Conchello and Seymour Knowles-Barley and Dongil Lee and Amelio Vázquez-Reina and Verena Kaynig and Thouis Raymond Jones and Mike Roberts and Josh Lyskowski Morgan and Juan Carlos Tapia and H. Sebastian Seung and William Gray Roncal and Joshua Tzvi Vogelstein and Randal Burns and Daniel Lewis Sussman and Carey Eldin Priebe and Hanspeter Pfister and Jeff William Lichtman"
}
@ARTICLE{2013arXiv1303.7186K,
author = {{Kaynig}, V. and {Vazquez-Reina}, A. and {Knowles-Barley}, S. and
{Roberts}, M. and {Jones}, T.~R. and {Kasthuri}, N. and {Miller}, E. and
{Lichtman}, J. and {Pfister}, H.},
title = "{Large-Scale Automatic Reconstruction of Neuronal Processes from Electron Microscopy Images}",
journal = {ArXiv e-prints},
archivePrefix = "arXiv",
eprint = {1303.7186},
primaryClass = "q-bio.NC",
keywords = {Quantitative Biology - Neurons and Cognition, Computer Science - Computer Vision and Pattern Recognition},
year = 2013,
month = mar,
adsurl = {http://adsabs.harvard.edu/abs/2013arXiv1303.7186K},
adsnote = {Provided by the SAO/NASA Astrophysics Data System}
}
@article{segEM,
author = "ManuelBerning and Kevin Boergens and Moritz Helmstaedter",
title = "SegEM: Efficient Image Analysis for High-Resolution Connectomics",
journal = "Neuron",
issn = "Volume 87, Issue 6, 23 September 2015, Pages 1193-1206"
}
@article{DBLP:journals/corr/AbadiBCCDDDGIIK16,
author = {Mart{\'{\i}}n Abadi and
Paul Barham and
Jianmin Chen and
Zhifeng Chen and
Andy Davis and
Jeffrey Dean and
Matthieu Devin and
Sanjay Ghemawat and
Geoffrey Irving and
Michael Isard and
Manjunath Kudlur and
Josh Levenberg and
Rajat Monga and
Sherry Moore and
Derek Gordon Murray and
Benoit Steiner and
Paul A. Tucker and
Vijay Vasudevan and
Pete Warden and
Martin Wicke and
Yuan Yu and
Xiaoqiang Zhang},
title = {TensorFlow: {A} system for large-scale machine learning},
journal = {CoRR},
volume = {abs/1605.08695},
year = {2016},
url = {http://arxiv.org/abs/1605.08695},
archivePrefix = {arXiv},
eprint = {1605.08695},
timestamp = {Wed, 07 Jun 2017 14:41:30 +0200},
biburl = {http://dblp.org/rec/bib/journals/corr/AbadiBCCDDDGIIK16},
bibsource = {dblp computer science bibliography, http://dblp.org}
}
@article{horovod,
title = "Meet Horovod: Uber’s Open Source Distributed Deep Learning Framework for TensorFlow",
author = "Alex Sergeev and Mike Del Balso",
url = "https://eng.uber.com/horovod/"
}
@inproceedings{40808,
title = {An Empirical study of learning rates in deep neural networks for speech recognition},
author = {Andrew Senior and Georg Heigold and Marc'aurelio Ranzato and Ke Yang},
year = {2013},
booktitle = {Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP)},
address = {Vancouver, CA}
}
@ARTICLE{10.3389/fnana.2015.00142,
AUTHOR={Arganda-Carreras, Ignacio and Turaga, Srinivas C. and Berger, Daniel R. and Cireşan, Dan and Giusti, Alessandro and Gambardella, Luca M. and Schmidhuber, JÃŒrgen and Laptev, Dmitry and Dwivedi, Sarvesh and Buhmann, Joachim M. and Liu, Ting and Seyedhosseini, Mojtaba and Tasdizen, Tolga and Kamentsky, Lee and Burget, Radim and Uher, Vaclav and Tan, Xiao and Sun, Changming and Pham, Tuan D. and Bas, Erhan and Uzunbas, Mustafa G. and Cardona, Albert and Schindelin, Johannes and Seung, H. Sebastian},
TITLE={Crowdsourcing the creation of image segmentation algorithms for connectomics},
JOURNAL={Frontiers in Neuroanatomy},
VOLUME={9},
PAGES={142},
YEAR={2015},
URL={https://www.frontiersin.org/article/10.3389/fnana.2015.00142},
DOI={10.3389/fnana.2015.00142},
ISSN={1662-5129},
ABSTRACT={To stimulate progress in automating the reconstruction of neural circuits,
we organized the first international challenge on 2D segmentation
of electron microscopic (EM) images of the brain. Participants submitted
boundary maps predicted for a test set of images, and were scored
based on their agreement with ground truth from human experts. The
winning team had no prior experience with EM images, and employed
a convolutional network. This ``deep learning'' approach has since
become accepted as a standard for segmentation of EM images. The challenge
has continued to accept submissions, and the best so far has resulted
from cooperation between two teams. The challenge has probably saturated,
as algorithms cannot progress beyond limits set by ambiguities inherent
in 2D scoring. Retrospective evaluation of the challenge scoring system
reveals that it was not sufficiently robust to variations in the widths
of neurite borders. We propose a solution to this problem, which should
be useful for a future 3D segmentation challenge.}
}
@ARTICLE{2014arXiv1412.6980K,
author = {{Kingma}, D.~P. and {Ba}, J.},
title = "{Adam: A Method for Stochastic Optimization}",
journal = {ArXiv e-prints},
archivePrefix = "arXiv",
eprint = {1412.6980},
primaryClass = "cs.LG",
keywords = {Computer Science - Learning},
year = 2014,
month = dec,
adsurl = {http://adsabs.harvard.edu/abs/2014arXiv1412.6980K},
adsnote = {Provided by the SAO/NASA Astrophysics Data System}
}
\relax
\citation{segEM}
\citation{KASTHURI2015648}
\citation{2013arXiv1303.7186K}
\citation{DBLP:journals/corr/RonnebergerFB15}
\citation{DBLP:journals/corr/CicekALBR16}
\citation{DBLP:journals/corr/JanuszewskiMLKD16}
\citation{Januszewski200675}
\citation{DBLP:journals/corr/AbadiBCCDDDGIIK16}
\citation{horovod}
\@writefile{toc}{\contentsline {section}{\numberline {I}Introduction}{1}}
\@writefile{lof}{\contentsline {figure}{\numberline {1}{\ignorespaces Flood filling network architecture}}{1}}
\newlabel{fig_ffn}{{1}{1}}
\@writefile{toc}{\contentsline {section}{\numberline {II}Background}{1}}
\@writefile{toc}{\contentsline {subsection}{\numberline {\unhbox \voidb@x \hbox {II-A}}Flood-filling network}{1}}
\citation{DBLP:journals/corr/JanuszewskiMLKD16}
\citation{Januszewski200675}
\citation{horovod}
\citation{DBLP:journals/corr/GoyalDGNWKTJH17}
\citation{DBLP:journals/corr/abs-1709-05011}
\citation{DBLP:journals/corr/abs-1708-02983}
\citation{40808}
\citation{2014arXiv1412.6980K}
\@writefile{toc}{\contentsline {subsection}{\numberline {\unhbox \voidb@x \hbox {II-B}}Distributed Training}{2}}
\@writefile{toc}{\contentsline {subsection}{\numberline {\unhbox \voidb@x \hbox {II-C}}Optimizers}{2}}
\@writefile{lof}{\contentsline {figure}{\numberline {2}{\ignorespaces Knights landing architecture.}}{2}}
\newlabel{fig_KNL}{{2}{2}}
\@writefile{toc}{\contentsline {subsection}{\numberline {\unhbox \voidb@x \hbox {II-D}}Intel Xeon Phi Knights Landing processor}{2}}
\citation{KASTHURI2015648}
\citation{10.3389/fnana.2015.00142}
\@writefile{toc}{\contentsline {section}{\numberline {III}Experiment details}{3}}
\@writefile{toc}{\contentsline {subsection}{\numberline {\unhbox \voidb@x \hbox {III-A}}Data}{3}}
\@writefile{toc}{\contentsline {subsection}{\numberline {\unhbox \voidb@x \hbox {III-B}}Network}{3}}
\@writefile{toc}{\contentsline {subsection}{\numberline {\unhbox \voidb@x \hbox {III-C}}Evaluation}{3}}
\@writefile{toc}{\contentsline {section}{\numberline {IV}Results}{3}}
\@writefile{toc}{\contentsline {subsection}{\numberline {\unhbox \voidb@x \hbox {IV-A}}Training and inference}{3}}
\@writefile{toc}{\contentsline {subsection}{\numberline {\unhbox \voidb@x \hbox {IV-B}}Visualization of network layers}{3}}
\@writefile{toc}{\contentsline {section}{\numberline {V}Conclusion}{3}}
\bibstyle{IEEEtran}
\bibdata{bib.bib}
\bibcite{segEM}{1}
\bibcite{KASTHURI2015648}{2}
\bibcite{2013arXiv1303.7186K}{3}
\bibcite{DBLP:journals/corr/RonnebergerFB15}{4}
\bibcite{DBLP:journals/corr/CicekALBR16}{5}
\@writefile{lof}{\contentsline {figure}{\numberline {3}{\ignorespaces slice of AC4 data (mouse neocortext): raw EM image (left), CLAHE preprocessed data (middle), human-labeled ground truth (right)}}{4}}
\@writefile{lof}{\contentsline {figure}{\numberline {4}{\ignorespaces slice of AC3 raw data (left), CLAHE preprocessed (middle), and ground truth (right)}}{4}}
\@writefile{lot}{\contentsline {table}{\numberline {I}{\ignorespaces Single-node benchmark results}}{4}}
\newlabel{table_inference_results}{{I}{4}}
\@writefile{toc}{\contentsline {section}{References}{4}}
\@writefile{lof}{\contentsline {figure}{\numberline {5}{\ignorespaces top: slice of ffn segmentation, allowing cell sizes bigger than 100 (left), 1k (middle), and 10k (right). bottom: corresponding probability object masks}}{5}}
\@writefile{lot}{\contentsline {table}{\numberline {II}{\ignorespaces Single-node benchmark results}}{5}}
\newlabel{table_benchmarks}{{II}{5}}
\@writefile{lof}{\contentsline {figure}{\numberline {6}{\ignorespaces Learning curve for increasing number of horovod MPI ranks: 256 (red), 512 (green), 1024 (grey). With increasing number of ranks, the time to fixed loss decreases.}}{5}}
\newlabel{fig_strong_scaling}{{6}{5}}
\@writefile{lof}{\contentsline {figure}{\numberline {7}{\ignorespaces Learning curve for training on AC4 without (orange) and with (blue) CLAHE preprocessing. Preprocessing helps accelerate the training process.}}{5}}
\newlabel{fig_impact_of_prepr}{{7}{5}}
\bibcite{DBLP:journals/corr/JanuszewskiMLKD16}{6}
\bibcite{Januszewski200675}{7}
\bibcite{DBLP:journals/corr/AbadiBCCDDDGIIK16}{8}
\bibcite{horovod}{9}
\bibcite{DBLP:journals/corr/GoyalDGNWKTJH17}{10}
\bibcite{DBLP:journals/corr/abs-1709-05011}{11}
\bibcite{DBLP:journals/corr/abs-1708-02983}{12}
\bibcite{40808}{13}
\bibcite{2014arXiv1412.6980K}{14}
\bibcite{10.3389/fnana.2015.00142}{15}
\@writefile{lof}{\contentsline {figure}{\numberline {8}{\ignorespaces histogram of cell sizes}}{6}}
\newlabel{figure_histogram_cell_sizes}{{8}{6}}
\@writefile{lof}{\contentsline {figure}{\numberline {9}{\ignorespaces pixel-based metrics during training}}{7}}
\newlabel{fig_training_metrics}{{9}{7}}
\@writefile{lof}{\contentsline {figure}{\numberline {10}{\ignorespaces ffn activations of a test image in first layer, highlighting edges, and in the final layer, showing sparse, localized features.}}{8}}
\newlabel{fig_ffn_activations}{{10}{8}}
\@writefile{lof}{\contentsline {figure}{\numberline {11}{\ignorespaces ffn weights in the first layer.}}{8}}
\newlabel{fig_ffn_weights}{{11}{8}}
% Generated by IEEEtran.bst, version: 1.14 (2015/08/26)
\begin{thebibliography}{10}
\providecommand{\url}[1]{#1}
\csname url@samestyle\endcsname
\providecommand{\newblock}{\relax}
\providecommand{\bibinfo}[2]{#2}
\providecommand{\BIBentrySTDinterwordspacing}{\spaceskip=0pt\relax}
\providecommand{\BIBentryALTinterwordstretchfactor}{4}
\providecommand{\BIBentryALTinterwordspacing}{\spaceskip=\fontdimen2\font plus
\BIBentryALTinterwordstretchfactor\fontdimen3\font minus
\fontdimen4\font\relax}
\providecommand{\BIBforeignlanguage}[2]{{%
\expandafter\ifx\csname l@#1\endcsname\relax
\typeout{** WARNING: IEEEtran.bst: No hyphenation pattern has been}%
\typeout{** loaded for the language `#1'. Using the pattern for}%
\typeout{** the default language instead.}%
\else
\language=\csname l@#1\endcsname
\fi
#2}}
\providecommand{\BIBdecl}{\relax}
\BIBdecl
\bibitem{segEM}
ManuelBerning, K.~Boergens, and M.~Helmstaedter, ``Segem: Efficient image
analysis for high-resolution connectomics,'' \emph{Neuron}.
\bibitem{KASTHURI2015648}
\BIBentryALTinterwordspacing
N.~Kasthuri, K.~Hayworth, D.~Berger, R.~Schalek, J.~Conchello,
S.~Knowles-Barley, D.~Lee, A.~Vázquez-Reina, V.~Kaynig, T.~Jones,
M.~Roberts, J.~Morgan, J.~Tapia, H.~Seung, W.~Roncal, J.~Vogelstein,
R.~Burns, D.~Sussman, C.~Priebe, H.~Pfister, and J.~Lichtman, ``Saturated
reconstruction of a volume of neocortex,'' \emph{Cell}, vol. 162, no.~3, pp.
648 -- 661, 2015. [Online]. Available:
\url{http://www.sciencedirect.com/science/article/pii/S0092867415008247}
\BIBentrySTDinterwordspacing
\bibitem{2013arXiv1303.7186K}
V.~{Kaynig}, A.~{Vazquez-Reina}, S.~{Knowles-Barley}, M.~{Roberts}, T.~R.
{Jones}, N.~{Kasthuri}, E.~{Miller}, J.~{Lichtman}, and H.~{Pfister},
``{Large-Scale Automatic Reconstruction of Neuronal Processes from Electron
Microscopy Images},'' \emph{ArXiv e-prints}, Mar. 2013.
\bibitem{DBLP:journals/corr/RonnebergerFB15}
\BIBentryALTinterwordspacing
O.~Ronneberger, P.~Fischer, and T.~Brox, ``U-net: Convolutional networks for
biomedical image segmentation,'' \emph{CoRR}, vol. abs/1505.04597, 2015.
[Online]. Available: \url{http://arxiv.org/abs/1505.04597}
\BIBentrySTDinterwordspacing
\bibitem{DBLP:journals/corr/CicekALBR16}
\BIBentryALTinterwordspacing
{\"{O}}.~{\c{C}}i{\c{c}}ek, A.~Abdulkadir, S.~S. Lienkamp, T.~Brox, and
O.~Ronneberger, ``3d u-net: Learning dense volumetric segmentation from
sparse annotation,'' \emph{CoRR}, vol. abs/1606.06650, 2016. [Online].
Available: \url{http://arxiv.org/abs/1606.06650}
\BIBentrySTDinterwordspacing
\bibitem{DBLP:journals/corr/JanuszewskiMLKD16}
\BIBentryALTinterwordspacing
M.~Januszewski, J.~Maitin{-}Shepard, P.~Li, J.~Kornfeld, W.~Denk, and V.~Jain,
``Flood-filling networks,'' \emph{CoRR}, vol. abs/1611.00421, 2016. [Online].
Available: \url{http://arxiv.org/abs/1611.00421}
\BIBentrySTDinterwordspacing
\bibitem{Januszewski200675}
\BIBentryALTinterwordspacing
M.~Januszewski, J.~Kornfeld, P.~H. Li, A.~Pope, T.~Blakely, L.~Lindsey, J.~B.
Maitin-Shepard, M.~Tyka, W.~Denk, and V.~Jain, ``High-precision automated
reconstruction of neurons with flood-filling networks,'' \emph{bioRxiv},
2017. [Online]. Available:
\url{https://www.biorxiv.org/content/early/2017/10/09/200675}
\BIBentrySTDinterwordspacing
\bibitem{DBLP:journals/corr/AbadiBCCDDDGIIK16}
\BIBentryALTinterwordspacing
M.~Abadi, P.~Barham, J.~Chen, Z.~Chen, A.~Davis, J.~Dean, M.~Devin,
S.~Ghemawat, G.~Irving, M.~Isard, M.~Kudlur, J.~Levenberg, R.~Monga,
S.~Moore, D.~G. Murray, B.~Steiner, P.~A. Tucker, V.~Vasudevan, P.~Warden,
M.~Wicke, Y.~Yu, and X.~Zhang, ``Tensorflow: {A} system for large-scale
machine learning,'' \emph{CoRR}, vol. abs/1605.08695, 2016. [Online].
Available: \url{http://arxiv.org/abs/1605.08695}
\BIBentrySTDinterwordspacing
\bibitem{horovod}
\BIBentryALTinterwordspacing
A.~Sergeev and M.~D. Balso, ``Meet horovod: Uber’s open source distributed
deep learning framework for tensorflow.'' [Online]. Available:
\url{https://eng.uber.com/horovod/}
\BIBentrySTDinterwordspacing
\bibitem{DBLP:journals/corr/GoyalDGNWKTJH17}
\BIBentryALTinterwordspacing
P.~Goyal, P.~Doll{\'{a}}r, R.~B. Girshick, P.~Noordhuis, L.~Wesolowski,
A.~Kyrola, A.~Tulloch, Y.~Jia, and K.~He, ``Accurate, large minibatch {SGD:}
training imagenet in 1 hour,'' \emph{CoRR}, vol. abs/1706.02677, 2017.
[Online]. Available: \url{http://arxiv.org/abs/1706.02677}
\BIBentrySTDinterwordspacing
\bibitem{DBLP:journals/corr/abs-1709-05011}
\BIBentryALTinterwordspacing
Y.~You, Z.~Zhang, C.~Hsieh, and J.~Demmel, ``100-epoch imagenet training with
alexnet in 24 minutes,'' \emph{CoRR}, vol. abs/1709.05011, 2017. [Online].
Available: \url{http://arxiv.org/abs/1709.05011}
\BIBentrySTDinterwordspacing
\bibitem{DBLP:journals/corr/abs-1708-02983}
\BIBentryALTinterwordspacing
Y.~You, A.~Bulu{\c{c}}, and J.~Demmel, ``Scaling deep learning on {GPU} and
knights landing clusters,'' \emph{CoRR}, vol. abs/1708.02983, 2017. [Online].
Available: \url{http://arxiv.org/abs/1708.02983}
\BIBentrySTDinterwordspacing
\bibitem{40808}
A.~Senior, G.~Heigold, M.~Ranzato, and K.~Yang, ``An empirical study of
learning rates in deep neural networks for speech recognition,'' in
\emph{Proceedings of the IEEE International Conference on Acoustics, Speech,
and Signal Processing (ICASSP)}, Vancouver, CA, 2013.
\bibitem{2014arXiv1412.6980K}
D.~P. {Kingma} and J.~{Ba}, ``{Adam: A Method for Stochastic Optimization},''
\emph{ArXiv e-prints}, Dec. 2014.
\bibitem{10.3389/fnana.2015.00142}
\BIBentryALTinterwordspacing
I.~Arganda-Carreras, S.~C. Turaga, D.~R. Berger, D.~Cireşan, A.~Giusti,
L.~M. Gambardella, J.~Schmidhuber, D.~Laptev, S.~Dwivedi, J.~M. Buhmann,
T.~Liu, M.~Seyedhosseini, T.~Tasdizen, L.~Kamentsky, R.~Burget, V.~Uher,
X.~Tan, C.~Sun, T.~D. Pham, E.~Bas, M.~G. Uzunbas, A.~Cardona, J.~Schindelin,
and H.~S. Seung, ``Crowdsourcing the creation of image segmentation
algorithms for connectomics,'' \emph{Frontiers in Neuroanatomy}, vol.~9, p.
142, 2015. [Online]. Available:
\url{https://www.frontiersin.org/article/10.3389/fnana.2015.00142}
\BIBentrySTDinterwordspacing
\end{thebibliography}
This is BibTeX, Version 0.99d (TeX Live 2017)
Capacity: max_strings=100000, hash_size=100000, hash_prime=85009
The top-level auxiliary file: deepseg.aux
The style file: IEEEtran.bst
Reallocated singl_function (elt_size=4) to 100 items from 50.
Reallocated singl_function (elt_size=4) to 100 items from 50.
Reallocated singl_function (elt_size=4) to 100 items from 50.
Reallocated wiz_functions (elt_size=4) to 6000 items from 3000.
Reallocated singl_function (elt_size=4) to 100 items from 50.
Database file #1: bib.bib.bib
-- IEEEtran.bst version 1.14 (2015/08/26) by Michael Shell.
-- http://www.michaelshell.org/tex/ieeetran/bibtex/
-- See the "IEEEtran_bst_HOWTO.pdf" manual for usage information.
Warning--empty year in segEM
Warning--empty journal in horovod
Warning--empty year in horovod
Done.
You've used 15 entries,
4087 wiz_defined-function locations,
913 strings with 12201 characters,
and the built_in function-call counts, 10777 in all, are:
= -- 739
> -- 532
< -- 21
+ -- 260
- -- 127
* -- 594
:= -- 1684
add.period$ -- 30
call.type$ -- 15
change.case$ -- 15
chr.to.int$ -- 27
cite$ -- 18
duplicate$ -- 799
empty$ -- 808
format.name$ -- 142
if$ -- 2414
int.to.chr$ -- 0
int.to.str$ -- 15
missing$ -- 201
newline$ -- 90
num.names$ -- 15
pop$ -- 536
preamble$ -- 1
purify$ -- 0
quote$ -- 2
skip$ -- 768
stack$ -- 0
substring$ -- 84
swap$ -- 585
text.length$ -- 12
text.prefix$ -- 0
top$ -- 5
type$ -- 15
warning$ -- 3
while$ -- 19
width$ -- 17
write$ -- 184
(There were 3 warnings)
# Fdb version 3
["bibtex deepseg"] 1525892734 "deepseg.aux" "deepseg.bbl" "deepseg" 1526063334
"/usr/local/texlive/2017/texmf-dist/bibtex/bst/IEEEtran/IEEEtran.bst" 1480098433 57748 7c8250ecf02814ce6ddc0cdbb63df1dd ""
"bib.bib" 1525461078 11640 e2983864f279eb41be33f9c6b6e807da ""
"deepseg.aux" 1526063333 5070 dd0274774eb82e6b9a97e922c5596e51 ""
(generated)
"deepseg.blg"
"deepseg.bbl"
["pdflatex"] 1526063330 "deepseg.tex" "deepseg.pdf" "deepseg" 1526063334
"/usr/local/texlive/2017/texmf-dist/fonts/enc/dvips/base/8r.enc" 1480098666 4850 80dc9bab7f31fb78a000ccfed0e27cab ""
"/usr/local/texlive/2017/texmf-dist/fonts/map/fontname/texfonts.map" 1480098670 3287 e6b82fe08f5336d4d5ebc73fb1152e87 ""
"/usr/local/texlive/2017/texmf-dist/fonts/tfm/adobe/times/ptmb7t.tfm" 1480098689 2172 fd0c924230362ff848a33632ed45dc23 ""
"/usr/local/texlive/2017/texmf-dist/fonts/tfm/adobe/times/ptmb8r.tfm" 1480098689 4524 6bce29db5bc272ba5f332261583fee9c ""
"/usr/local/texlive/2017/texmf-dist/fonts/tfm/adobe/times/ptmbi7t.tfm" 1480098689 2228 e564491c42a4540b5ebb710a75ff306c ""
"/usr/local/texlive/2017/texmf-dist/fonts/tfm/adobe/times/ptmbi8r.tfm" 1480098689 4480 10409ed8bab5aea9ec9a78028b763919 ""
"/usr/local/texlive/2017/texmf-dist/fonts/tfm/adobe/times/ptmr7t.tfm" 1480098689 2124 2601a75482e9426d33db523edf23570a ""
"/usr/local/texlive/2017/texmf-dist/fonts/tfm/adobe/times/ptmr8r.tfm" 1480098689 4408 25b74d011a4c66b7f212c0cc3c90061b ""
"/usr/local/texlive/2017/texmf-dist/fonts/tfm/adobe/times/ptmrc7t.tfm" 1480098689 2680 312a2d12b1f1df8ee0212e7ba1962402 ""
"/usr/local/texlive/2017/texmf-dist/fonts/tfm/adobe/times/ptmri7t.tfm" 1480098689 2288 f478fc8fed18759effb59f3dad7f3084 ""
"/usr/local/texlive/2017/texmf-dist/fonts/tfm/adobe/times/ptmri8r.tfm" 1480098689 4640 532ca3305aad10cc01d769f3f91f1029 ""
"/usr/local/texlive/2017/texmf-dist/fonts/tfm/public/cm/cmmi6.tfm" 1480098701 1512 f21f83efb36853c0b70002322c1ab3ad ""
"/usr/local/texlive/2017/texmf-dist/fonts/tfm/public/cm/cmmi8.tfm" 1480098701 1520 eccf95517727cb11801f4f1aee3a21b4 ""
"/usr/local/texlive/2017/texmf-dist/fonts/tfm/public/cm/cmr6.tfm" 1480098701 1300 b62933e007d01cfd073f79b963c01526 ""
"/usr/local/texlive/2017/texmf-dist/fonts/tfm/public/cm/cmr8.tfm" 1480098701 1292 21c1c5bfeaebccffdb478fd231a0997d ""
"/usr/local/texlive/2017/texmf-dist/fonts/tfm/public/cm/cmsy6.tfm" 1480098701 1116 933a60c408fc0a863a92debe84b2d294 ""
"/usr/local/texlive/2017/texmf-dist/fonts/tfm/public/cm/cmsy8.tfm" 1480098701 1120 8b7d695260f3cff42e636090a8002094 ""
"/usr/local/texlive/2017/texmf-dist/fonts/type1/public/amsfonts/cm/cmmi10.pfb" 1480098733 36299 5f9df58c2139e7edcf37c8fca4bd384d ""
"/usr/local/texlive/2017/texmf-dist/fonts/type1/public/amsfonts/cm/cmmi7.pfb" 1480098733 36281 c355509802a035cadc5f15869451dcee ""
"/usr/local/texlive/2017/texmf-dist/fonts/type1/public/amsfonts/cm/cmr10.pfb" 1480098733 35752 024fb6c41858982481f6968b5fc26508 ""
"/usr/local/texlive/2017/texmf-dist/fonts/type1/public/amsfonts/cm/cmr7.pfb" 1480098733 32762 224316ccc9ad3ca0423a14971cfa7fc1 ""
"/usr/local/texlive/2017/texmf-dist/fonts/type1/public/amsfonts/cm/cmsy10.pfb" 1480098733 32569 5e5ddc8df908dea60932f3c484a54c0d ""
"/usr/local/texlive/2017/texmf-dist/fonts/type1/public/amsfonts/cm/cmsy7.pfb" 1480098733 32716 08e384dc442464e7285e891af9f45947 ""
"/usr/local/texlive/2017/texmf-dist/fonts/type1/urw/times/utmb8a.pfb" 1480098746 44729 811d6c62865936705a31c797a1d5dada ""
"/usr/local/texlive/2017/texmf-dist/fonts/type1/urw/times/utmbi8a.pfb" 1480098746 44656 0cbca70e0534538582128f6b54593cca ""
"/usr/local/texlive/2017/texmf-dist/fonts/type1/urw/times/utmr8a.pfb" 1480098746 46026 6dab18b61c907687b520c72847215a68 ""
"/usr/local/texlive/2017/texmf-dist/fonts/type1/urw/times/utmri8a.pfb" 1480098746 45458 a3faba884469519614ca56ba5f6b1de1 ""
"/usr/local/texlive/2017/texmf-dist/fonts/vf/adobe/times/ptmb7t.vf" 1480098758 1372 788387fea833ef5963f4c5bffe33eb89 ""
"/usr/local/texlive/2017/texmf-dist/fonts/vf/adobe/times/ptmbi7t.vf" 1480098758 1384 6ac0f8b839230f5d9389287365b243c0 ""
"/usr/local/texlive/2017/texmf-dist/fonts/vf/adobe/times/ptmr7t.vf" 1480098758 1380 0ea3a3370054be6da6acd929ec569f06 ""
"/usr/local/texlive/2017/texmf-dist/fonts/vf/adobe/times/ptmrc7t.vf" 1480098758 1948 7330aeef3af211edff3b35fb2c12a0fd ""
"/usr/local/texlive/2017/texmf-dist/fonts/vf/adobe/times/ptmri7t.vf" 1480098758 1384 a9d8adaf491ce34e5fba99dc7bbe5f39 ""
"/usr/local/texlive/2017/texmf-dist/tex/context/base/mkii/supp-pdf.mkii" 1480098806 71627 94eb9990bed73c364d7f53f960cc8c5b ""
"/usr/local/texlive/2017/texmf-dist/tex/generic/oberdiek/etexcmds.sty" 1480098815 7612 729a8cc22a1ee0029997c7f74717ae05 ""
"/usr/local/texlive/2017/texmf-dist/tex/generic/oberdiek/ifluatex.sty" 1480098815 7324 2310d1247db0114eb4726807c8837a0e ""
"/usr/local/texlive/2017/texmf-dist/tex/generic/oberdiek/ifpdf.sty" 1490564930 1251 d170e11a3246c3392bc7f59595af42cb ""
"/usr/local/texlive/2017/texmf-dist/tex/generic/oberdiek/infwarerr.sty" 1480098815 8253 473e0e41f9adadb1977e8631b8f72ea6 ""
"/usr/local/texlive/2017/texmf-dist/tex/generic/oberdiek/kvdefinekeys.sty" 1480098815 5152 b67a3a964ad9851e095110c854a1d461 ""
"/usr/local/texlive/2017/texmf-dist/tex/generic/oberdiek/kvsetkeys.sty" 1480098815 14040 ac8866aac45982ac84021584b0abb252 ""
"/usr/local/texlive/2017/texmf-dist/tex/generic/oberdiek/ltxcmds.sty" 1480098815 18425 5b3c0c59d76fac78978b5558e83c1f36 ""
"/usr/local/texlive/2017/texmf-dist/tex/generic/oberdiek/pdftexcmds.sty" 1490564930 20151 72b3c7cacb61f7dd527505c39a23f7c1 ""
"/usr/local/texlive/2017/texmf-dist/tex/latex/graphics-cfg/graphics.cfg" 1480098830 1224 978390e9c2234eab29404bc21b268d1e ""
"/usr/local/texlive/2017/texmf-dist/tex/latex/graphics-def/pdftex.def" 1485129666 58250 3792a9d2d1d664ee8c742498e295b051 ""
"/usr/local/texlive/2017/texmf-dist/tex/latex/graphics/graphics.sty" 1492297155 14603 b288c52bd5d46d593af31dbc7e548236 ""
"/usr/local/texlive/2017/texmf-dist/tex/latex/graphics/graphicx.sty" 1480098830 8125 557ab9f1bfa80d369fb45a914aa8a3b4 ""
"/usr/local/texlive/2017/texmf-dist/tex/latex/graphics/keyval.sty" 1480098830 2594 d18d5e19aa8239cf867fa670c556d2e9 ""
"/usr/local/texlive/2017/texmf-dist/tex/latex/graphics/trig.sty" 1480098830 3980 0a268fbfda01e381fa95821ab13b6aee ""
"/usr/local/texlive/2017/texmf-dist/tex/latex/latexconfig/epstopdf-sys.cfg" 1480098833 678 4792914a8f45be57bb98413425e4c7af ""
"/usr/local/texlive/2017/texmf-dist/tex/latex/oberdiek/epstopdf-base.sty" 1480098836 12095 5337833c991d80788a43d3ce26bd1c46 ""
"/usr/local/texlive/2017/texmf-dist/tex/latex/oberdiek/grfext.sty" 1480098836 7075 2fe3d848bba95f139de11ded085e74aa ""
"/usr/local/texlive/2017/texmf-dist/tex/latex/oberdiek/kvoptions.sty" 1480098836 22417 1d9df1eb66848aa31b18a593099cf45c ""
"/usr/local/texlive/2017/texmf-dist/tex/latex/psnfss/ot1ptm.fd" 1480098837 961 15056f4a61917ceed3a44e4ac11fcc52 ""
"/usr/local/texlive/2017/texmf-dist/web2c/texmf.cnf" 1494087824 32646 eadc4ca26cdbe7105ac7c593aa8c4f72 ""
"/usr/local/texlive/2017/texmf-var/fonts/map/pdftex/updmap/pdftex.map" 1495593472 2350277 a699055bee05bf8a40b0504752487295 ""
"/usr/local/texlive/2017/texmf-var/web2c/pdftex/pdflatex.fmt" 1495593487 4129592 0e6cfee100d68e6d96b92c0258f2ff73 ""
"/usr/local/texlive/2017/texmf.cnf" 1495593465 577 2b71d4d888f9e5560b2e99985915a9fa ""
"IEEEtran.cls" 1517282870 281957 5b2e4fa15b0f7eabb840ebf67df4c0f7 ""
"ac3_clahe.png" 1525379162 640621 2a9a2ef7def14e3e26ec1b2ed978d2d6 ""
"ac3_gt.png" 1525379179 36285 70c6be444fed07aebd698e971cedd818 ""
"ac3_model.png" 1525379123 73716 11ec2a1504a22c7d62311eb90f33a3d5 ""
"ac3_model_100.png" 1525379152 79099 b1ee5b43228531e8a358882202a5b224 ""
"ac3_model_10k.png" 1525379143 57720 1b2dd4c15b1c2f88d0f929fc44c5cc30 ""
"ac3_raw.png" 1525379172 483722 f068ae420a7b1f119217a6f0eff0e7fe ""
"ac4_clahe.png" 1525379100 597032 91ed2f37d51358d48202033c8740511e ""
"ac4_gt.png" 1525379113 38503 6f7311a3e87dfdf0474e5661cf3ae294 ""
"ac4_raw.png" 1525378985 491557 ca7f3fb9a74f8c58dbd6a7ac37530cd6 ""
"cell_sizes.png" 1525365144 26009 9298dbf550d3945e1d176e2e919d9b14 ""