Skip to content

Commit

Permalink
All done
Browse files Browse the repository at this point in the history
  • Loading branch information
kumar-shridhar committed Dec 4, 2018
1 parent 5dc2312 commit 5fd4010
Showing 1 changed file with 53 additions and 0 deletions.
53 changes: 53 additions & 0 deletions Chapter6/chapter6.tex
Original file line number Diff line number Diff line change
Expand Up @@ -100,6 +100,59 @@ \section{BayesCNN for Generative Adversarial Networks}

\subsection{Our approach}

We based our work on the paper : Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks by Radford \cite{DBLP:journals/corr/RadfordMC15}. We used the architecture of a deep convolutional generative adversarial networks (DCGANs) that learns a hierarchy of representations from object parts to scenes in both the generator and discriminator.
The generator used in the Network is shown in Table \ref{tab:GeneratorArchitecture}. The architecture is kept similar to the architecture used in DCGAN paper \cite{DBLP:journals/corr/RadfordMC15}. Table \ref{tab:DiscriminatorArchitecture} shows the discriminator network with Bayesian Convolutional Layers.

\begin{table}[H]
\centering
\renewcommand{\arraystretch}{2}
\begin{tabular}{c c c c c}
\hline
layer type & width & stride & padding & nonlinearity \\ [0.5ex]
\hline
ConvolutionTranspose ($4\times4$) & ngf * 8 & 1 & 0 & ReLU \\


ConvolutionTranspose ($4\times4$) & ngf * 4 & 2 & 1 & ReLU \\


ConvolutionTranspose ($4\times4$) & ngf * 2 & 2 & 1 & ReLU \\

ConvolutionTranspose ($4\times4$) & ngf & 2 & 1 & ReLU \\

ConvolutionTranspose ($4\times4$) & nc & 2 & 1 & TanH \\ [1ex]
\hline
\end{tabular}
\renewcommand{\arraystretch}{1}
\caption{Generator architecture as defined in the paper. \cite{DBLP:journals/corr/RadfordMC15}}
\label{tab:GeneratorArchitecture}
\end{table}

\begin{table}[H]
\centering
\renewcommand{\arraystretch}{2}
\begin{tabular}{c c c c c}
\hline
layer type & width & stride & padding & nonlinearity \\ [0.5ex]
\hline
Convolution ($4\times4$) & ndf & 2 & 1 & LeakyReLU \\


Convolution($4\times4$) & ndf * 2 & 2 & 1 & LeakyReLU \\


Convolution ($4\times4$) & ndf * 4 & 2 & 1 & LeakyReLU \\

Convolution ($4\times4$) & ndf * 8 & 2 & 1 & leakyReLU \\

ConvolutionTranspose ($4\times4$) & 1 & 1 & 0 & Sigmoid \\ [1ex]
\hline
\end{tabular}
\renewcommand{\arraystretch}{1}
\caption{Discriminator architecture with Bayesian Convolutional layers}
\label{tab:DiscriminatorArchitecture}
\end{table}


\subsection{Empirical Analysis}

Expand Down

0 comments on commit 5fd4010

Please sign in to comment.