Table 1 Layers of classifier neural network

0

Convolutional

M⁽⁰⁾=d, T⁽⁰⁾=T,

S⁽⁰⁾=(⌊T/6⌋+1)·2+1

1

Max pooling

M⁽¹⁾=M⁽⁰⁾·6, T⁽¹⁾=T⁽⁰⁾,

R⁽¹⁾=3 if applicable, else R⁽¹⁾=1

2

Convolutional

M⁽²⁾=M⁽¹⁾, T⁽²⁾=⌈T⁽¹⁾/R⁽¹⁾⌉,

S⁽²⁾=(⌊T/12⌋+1)·2+1

M⁽³⁾=M⁽²⁾·3, T⁽³⁾=T⁽²⁾

3

Fully connected

N⁽³⁾=M⁽³⁾·T⁽³⁾

4

Fully connected

\(N^{(4)} = \lfloor \sqrt {N^{(3)} \cdot N^{(5)}} \rfloor \)

5

Output

N⁽⁵⁾=n