1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142

//! Applies the nonlinear TanH function.
//!
//! Non-linearity activation function: y = sinh(x) / cosh(x)
//!
//! You might consider using ReLU as an alternative.
//!
//! ReLU, compared to TanH:
//!
//! * reduces the likelyhood of vanishing gradients
//! * increases the likelyhood of a more beneficial sparse representation
//! * can be computed faster
//! * is therefore the most popular activation function in DNNs as of this writing (2016).
use co::{IBackend, SharedTensor};
use conn;
use layer::*;
use util::ArcLock;

#[derive(Debug, Clone)]
#[allow(missing_copy_implementations)]
/// TanH Activation Layer
pub struct TanH;

//
// Tanh + TanhPointwise
// Only on CUDA
#[cfg(all(feature="cuda", not(feature="native")))]
impl<B: IBackend + conn::Tanh<f32> + conn::TanhPointwise<f32>> ILayer<B> for TanH {
    impl_ilayer_activation!();

    fn compute_in_place(&self) -> bool {
        true
    }

    fn reshape(&mut self,
               backend: ::std::rc::Rc<B>,
               input_data: &mut Vec<ArcLock<SharedTensor<f32>>>,
               input_gradient: &mut Vec<ArcLock<SharedTensor<f32>>>,
               weights_data: &mut Vec<ArcLock<SharedTensor<f32>>>,
               weights_gradient: &mut Vec<ArcLock<SharedTensor<f32>>>,
               output_data: &mut Vec<ArcLock<SharedTensor<f32>>>,
               output_gradient: &mut Vec<ArcLock<SharedTensor<f32>>>) {
        if let Some(inp) = input_data.get(0) {
            let read_inp = inp.read().unwrap();
            let input_desc = read_inp.desc();
            input_gradient[0].write().unwrap().resize(input_desc).unwrap();
            output_data[0].write().unwrap().resize(input_desc).unwrap();
            output_gradient[0].write().unwrap().resize(input_desc).unwrap();
        }
    }
}

#[cfg(all(feature="cuda", not(feature="native")))]
impl<B: IBackend + conn::Tanh<f32> + conn::TanhPointwise<f32>> ComputeOutput<f32, B> for TanH {
    fn compute_output(&self,
                      backend: &B,
                      _weights: &[&SharedTensor<f32>],
                      input_data: &[&SharedTensor<f32>],
                      output_data: &mut [&mut SharedTensor<f32>]) {
        match input_data.get(0) {
            Some(input) => backend.tanh_plain(input, output_data[0]).unwrap(),
            None => backend.tanh_pointwise_plain(output_data[0]).unwrap(),
        }
    }
}

#[cfg(all(feature="cuda", not(feature="native")))]
impl<B: IBackend + conn::Tanh<f32> + conn::TanhPointwise<f32>> ComputeInputGradient<f32, B> for TanH {
    fn compute_input_gradient(&self,
                              backend: &B,
                              weights_data: &[&SharedTensor<f32>],
                              output_data: &[&SharedTensor<f32>],
                              output_gradients: &[&SharedTensor<f32>],
                              input_data: &[&SharedTensor<f32>],
                              input_gradients: &mut [&mut SharedTensor<f32>]) {
        match output_data.get(0) {
            Some(_) => backend.tanh_grad_plain(output_data[0], output_gradients[0], input_data[0], input_gradients[0]).unwrap(),
            None => backend.tanh_pointwise_grad_plain(input_data[0], input_gradients[0]).unwrap(),
        }
    }
}

#[cfg(all(feature="cuda", not(feature="native")))]
impl<B: IBackend + conn::Tanh<f32> + conn::TanhPointwise<f32>> ComputeParametersGradient<f32, B> for TanH {}

//
// Tanh without TanhPointwise
// Only on CUDA
//
#[cfg(feature="native")]
impl<B: IBackend + conn::Tanh<f32>> ILayer<B> for TanH {
    impl_ilayer_activation!();

    fn reshape(&mut self,
               backend: ::std::rc::Rc<B>,
               input_data: &mut Vec<ArcLock<SharedTensor<f32>>>,
               input_gradient: &mut Vec<ArcLock<SharedTensor<f32>>>,
               weights_data: &mut Vec<ArcLock<SharedTensor<f32>>>,
               weights_gradient: &mut Vec<ArcLock<SharedTensor<f32>>>,
               output_data: &mut Vec<ArcLock<SharedTensor<f32>>>,
               output_gradient: &mut Vec<ArcLock<SharedTensor<f32>>>) {
        if let Some(inp) = input_data.get(0) {
            let read_inp = inp.read().unwrap();
            let input_desc = read_inp.desc();
            input_gradient[0].write().unwrap().resize(input_desc).unwrap();
            output_data[0].write().unwrap().resize(input_desc).unwrap();
            output_gradient[0].write().unwrap().resize(input_desc).unwrap();
        }
    }
}

#[cfg(feature="native")]
impl<B: IBackend + conn::Tanh<f32>> ComputeOutput<f32, B> for TanH {
    fn compute_output(&self,
                      backend: &B,
                      _weights: &[&SharedTensor<f32>],
                      input_data: &[&SharedTensor<f32>],
                      output_data: &mut [&mut SharedTensor<f32>]) {
        match input_data.get(0) {
            Some(input) => backend.tanh_plain(input, output_data[0]).unwrap(),
            None => panic!("No input provided for TanH layer."),
        }
    }
}

#[cfg(feature="native")]
impl<B: IBackend + conn::Tanh<f32>> ComputeInputGradient<f32, B> for TanH {
    fn compute_input_gradient(&self,
                              backend: &B,
                              weights_data: &[&SharedTensor<f32>],
                              output_data: &[&SharedTensor<f32>],
                              output_gradients: &[&SharedTensor<f32>],
                              input_data: &[&SharedTensor<f32>],
                              input_gradients: &mut [&mut SharedTensor<f32>]) {
        match output_data.get(0) {
            Some(_) => backend.tanh_grad_plain(output_data[0], output_gradients[0], input_data[0], input_gradients[0]).unwrap(),
            None => panic!("No output_data provided for TanH layer backward."),
        }
    }
}

#[cfg(feature="native")]
impl<B: IBackend + conn::Tanh<f32>> ComputeParametersGradient<f32, B> for TanH {}