Report TODO list

[miccio-dk]

1. Introduction C
2. Related work R
   1. Classical Approaches C
   2. Deep Learning R
   3. State of the art R
3. Methodology R
   1. Deep Learning C
      1. Multilayer Perceptron R
      2. Activation R
      3. Optimization R
      4. Convolutional Neural Networks (CNN) C
      5. Recurrent Neural Networks (RNN) R
      6. Autoencoders (AE) C
      7. Batch Normalization C
      8. Dropout C
   2. Data Representation C
      1. Time Domain Representation C
      2. Discrete Fourier Transform (DFT) C
      3. Short-Time Fourier Transform (STFT) C
   3. Processing R
      1. Dynamic Range Compression R
      2. Real/Imaginary R
   4. Evaluation C
      1. Signal-to-Noise Ratio (SNR) C
      2. Source Evaluation C
      3. STOI C
      4. PESQ C
4. Experimental Setup R
   1. Dataset R
      1. Pairs R
      2. Processing R
   2. Framework R
      1. Training R
      2. Results R
      3. Denoising R
   3. Models R
      1. Convolutional Autoencoder R
      2. Convolutional-Recurrent Model (+ Baseline) R
      3. Temporal Convolutional Network C
   4. Experiments R
      1. Choosing processing R
      2. Training models C
      3. Fine-tuning R
5. Experimental results R
   1. Observations on choice of processing R
      1. Impact of data processing R
      2. Impact of loss function window R
   2. Observations on preliminary training C
      1. Pink noise vs real-world noise C
      2. Magnitude vs logarithmic representation C
   3. Observations on fine-tuning ``
      1. Noise stationarity ``
      2. Impact of network depth RC
6. Conclusions R
   1. Synthesis R
   2. Limitations RC
   3. Outline RC

[miccio-dk]

Figure conventions:

• In python:

plt.figure(figsize=(16, 4), dpi=300)

• In LaTeX (fig_name should contain a reference to a chapter/section, e.g. proc_spec1):

\begin{figure}[h]
    \centering
    \includegraphics[width=0.75\textwidth]{figures/fig_name}
    \caption{caption text, starting in lowercase}
    \label{fig:fig_name}
\end{figure}