Table 2
From: Extreme learning machine versus classical feedforward network
Algorithm 2 LRF-ELM training procedure | |
|---|---|
Input: Training data, | |
Output: Trained LRF-ELM model | |
1: Randomly initialise convolutional filter weights | |
2: Orthogonalise filter weights | |
3: Calculate a matrix of hidden features \(\mathbf {H}\) (result of convolution and pooling) | |
4: Compute optimal weights for classification layer using Eq. 11 | |
5: return LRF-ELM (with filters set in Step 2 and output weights from Step 4) |