k.ladutenko/deepmie

Keine Beschreibung

Ravi Hegde 9961b26ed9 mxmodel		vor 6 Jahren
.ipynb_checkpoints	9961b26ed9 mxmodel	vor 6 Jahren
.mypy_cache	b2e2f34fad further dev	vor 6 Jahren
__pycache__	9961b26ed9 mxmodel	vor 6 Jahren
datasets	9961b26ed9 mxmodel	vor 6 Jahren
materials	1d7b8580a1 again	vor 6 Jahren
models	25d68667cb commit with assisted DE	vor 6 Jahren
results	65106a91a8 added rugate functions	vor 6 Jahren
README	e9a845a293 README created online with Bitbucket	vor 6 Jahren
Untitled.ipynb	a186f56b20 simple qx working code	vor 6 Jahren
Untitled1.ipynb	474d2ae715 settled on 2 mat designs	vor 6 Jahren
Untitled2.ipynb	474d2ae715 settled on 2 mat designs	vor 6 Jahren
Untitled3.ipynb	9961b26ed9 mxmodel	vor 6 Jahren
barplotting.ipynb	b2e2f34fad further dev	vor 6 Jahren
dataset_maker.ipynb	9961b26ed9 mxmodel	vor 6 Jahren
de2.py	474d2ae715 settled on 2 mat designs	vor 6 Jahren
de_conv.ipynb	65106a91a8 added rugate functions	vor 6 Jahren
hypoptim.ipynb	b2e2f34fad further dev	vor 6 Jahren
hyptune.py	1e493c3bd5 further develp in conv scatter nets	vor 6 Jahren
loss_defs.py	b2e2f34fad further dev	vor 6 Jahren
make_dataset.py	1d7b8580a1 again	vor 6 Jahren
makeqx.py	474d2ae715 settled on 2 mat designs	vor 6 Jahren
mtmm.pyx	a186f56b20 simple qx working code	vor 6 Jahren
mxmodel.ipynb	25d68667cb commit with assisted DE	vor 6 Jahren
normalized.ipynb	474d2ae715 settled on 2 mat designs	vor 6 Jahren
oldqx.py	9961b26ed9 mxmodel	vor 6 Jahren
pygmoimp.ipynb	25d68667cb commit with assisted DE	vor 6 Jahren
pymietest.ipynb	65106a91a8 added rugate functions	vor 6 Jahren
qxnew.py	474d2ae715 settled on 2 mat designs	vor 6 Jahren
qxplots.py	a186f56b20 simple qx working code	vor 6 Jahren
reinforcement.ipynb	474d2ae715 settled on 2 mat designs	vor 6 Jahren
ru2.ipynb	9961b26ed9 mxmodel	vor 6 Jahren
rugate_d.ipynb	474d2ae715 settled on 2 mat designs	vor 6 Jahren
scatternet.ipynb	9961b26ed9 mxmodel	vor 6 Jahren
scipydiffe.ipynb	65106a91a8 added rugate functions	vor 6 Jahren
scnets.py	9961b26ed9 mxmodel	vor 6 Jahren
setup.py	a186f56b20 simple qx working code	vor 6 Jahren
siti_fully-0000.params	9961b26ed9 mxmodel	vor 6 Jahren
siti_fully-symbol.json	9961b26ed9 mxmodel	vor 6 Jahren
smooth_x.ipynb	65106a91a8 added rugate functions	vor 6 Jahren
snlay.py	b2e2f34fad further dev	vor 6 Jahren
stack_testing.ipynb	9961b26ed9 mxmodel	vor 6 Jahren

		
				README
			
				https://bitbucket.org/rshegde/deepmie-deep-learning-electromagnetic-scattering

Abstract
===============================================================
We propose the use of Deep Convolutional Neural Networks (DCNN)
  in concert with Differential Evolution for solving inverse problems in
  electromagnetics.  We show that networks built using innovations like
  convolutional layers, batch normalization, parametric REctified Linear Unit
  (ReLU) and residual blocks achieve comparable performance to previously
  reported dense fully-connected networks but achieve drastic reduction in
  parameter weights and training epoch lengths. The proposed lightweight CNNs
  are used in concert with Differential Evolution global optimization to achieve
  8x speedup in convergence time. Systematic hyperparameter grid searches are
  reported to permit faster model discovery for application to other problems.

See the following arxiv paper for details

Requirements
================================================================
1) CUDA capable GPU 
2) Python 3.5
3) Numpy, Matplotlib, Pandas, Scipy, Scikitlearn
4) Cudatooklit 9.0, CuDNN 7.1
5) Mxnet-GPU, Mxnet-cumkl
6) Keras, Keras-mxnet
7) Scattnlay -- needed if you want to generate other datasets
8) Jupyter notebook 
9) ipyparallel for calling multiple runs at the same time

Getting started
==================================
Most of the code is placed in Jupyter notebooks for interactivity. 
PyMIEtest.ipynb   --  familiarizes you with generating datasets
Scatternets.ipynb -- allows building and visualizing models
hypoptim.ipynb -- hyperparameter grid search using scikit CV grid search
scipydiffe -- differential evolution using compiled models.