FMRSS Net: Fast Matrix Representation-Based Spectral-Spatial Feature Learning Convolutional Neural Network for Hyperspectral Image Classification

Table 1

---

Parameter Settings

---

INPUT:

(i) Input data:

CONVOLUTION 1:

(i) 6 feature maps

(ii) Kernel size: , where , rounds the elements to the nearest integers towards infinity.

(iii) Feature map size: ()

(iv) Training parameters number:

(v) Activation function: Sigmoid

MAX POOLING:

(i) 6 feature maps

(ii) Kernel size: , is set between

(iii) Feature map size: ()

(iv) Training parameters number:

(v) Pooling function: Max pooling sampling

CONVOLUTION 2:

(i) 12 feature maps

(ii) Kernel size: .

(iii) Feature map size: ()

(iv) Training parameters number:

(v) Activation function: Sigmoid

FULL CONNECTION:

(i) Units number:

(ii) Training parameters number:

(iii) Activation function: Softmax

OUTPUT:

(i) Units number , also named classes number

(ii) Training parameters number:

---