Table 2 Variables and their descriptions.
From: Hybrid deep learning framework for accurate classification of high dimensional genomic data
Variable | Description |
|---|---|
X | Raw genomic data matrix with n samples and d features |
Y | True class labels for each sample |
\(\theta _T\) | Parameters of the TabNet feature selector |
\(\theta _C\) | Parameters of the CNN classifier |
\(W_r\) | Learnable weight vector for feature refinement |
\(\widetilde{W}_r\) | Normalized refinement weights after softmax |
F | Selected feature matrix from TabNet (size \(n\times k\)) |
\(F'\) | Refined feature matrix after applying \(W_r\) |
\(C_i\) | i-th convolutional filter in the CNN layers |
\(A_i\) | Activated output of convolution \(C_i\) after ReLU |
\(P_i\) | Pooled output after applying max-pool to \(A_i\) |
V | Flattened feature vector formed from all pooled maps |
P | Prediction logits or probabilities from the CNN |
\(\hat{Y}\) | Final predicted labels |
\(A_T\) | Attention scores from TabNet (for interpretability) |
\(A_C\) | Activation maps from CNN (for interpretability) |
