学习笔记: 源码 filler.hpp 简记

1. GetFiller()

//通过调用GetFiller 函数，进行相应的继承类对象的创建。
template <typename Dtype>
Filler<Dtype>* GetFiller(const FillerParameter& param) { 
  const std::string& type = param.type();
  if (type == "constant") {
    return new ConstantFiller<Dtype>(param);
  } else if (type == "gaussian") {
    return new GaussianFiller<Dtype>(param);
  } else if (type == "positive_unitball") {
    return new PositiveUnitballFiller<Dtype>(param);
  } else if (type == "uniform") {
    return new UniformFiller<Dtype>(param);
  } else if (type == "xavier") {
    return new XavierFiller<Dtype>(param);
  } else if (type == "msra") {
    return new MSRAFiller<Dtype>(param);
  } else if (type == "bilinear") {
    return new BilinearFiller<Dtype>(param);
  } else {
    CHECK(false) << "Unknown filler name: " << param.type();
  }
  return (Filler<Dtype>*)(NULL);
}

2. Filler::Fill()

template <typename Dtype>
class Filler {
 public:
  explicit Filler(const FillerParameter& param) : filler_param_(param) {}
  virtual ~Filler() {}
  virtual void Fill(Blob<Dtype>* blob) = 0; //可以看出, Fill 函数将对一个blob 的数值进行初始化赋值
 protected:
  FillerParameter filler_param_;
};  // class Filler

3. 几个常见的参数初始化方式

// 常数初始化
/// @brief Fills a Blob with constant values @f$ x = 0 @f$.
template <typename Dtype>
class ConstantFiller : public Filler<Dtype> {
 public:
  explicit ConstantFiller(const FillerParameter& param)
      : Filler<Dtype>(param) {}
  virtual void Fill(Blob<Dtype>* blob) {
    Dtype* data = blob->mutable_cpu_data();
    const int count = blob->count();
    const Dtype value = this->filler_param_.value();
    CHECK(count);
    for (int i = 0; i < count; ++i) {
      data[i] = value;  //将所有的参数初始化为常数value
    }
    CHECK_EQ(this->filler_param_.sparse(), -1)
         << "Sparsity not supported by this Filler.";
  }
};

/// @brief Fills a Blob with uniformly distributed values @f$ x\sim U(a, b) @f$.
template <typename Dtype>
class UniformFiller : public Filler<Dtype> {
 public:
  explicit UniformFiller(const FillerParameter& param)
      : Filler<Dtype>(param) {}
  virtual void Fill(Blob<Dtype>* blob) {
    CHECK(blob->count());
    caffe_rng_uniform<Dtype>(blob->count(), Dtype(this->filler_param_.min()),//产生均匀分布的随机数初始化
        Dtype(this->filler_param_.max()), blob->mutable_cpu_data());   //默认的均匀分布为[0, 1]
    CHECK_EQ(this->filler_param_.sparse(), -1)
         << "Sparsity not supported by this Filler.";
  }
};

/// @brief Fills a Blob with Gaussian-distributed values @f$ x = a @f$.
template <typename Dtype>
class GaussianFiller : public Filler<Dtype> {
 public:
  explicit GaussianFiller(const FillerParameter& param)
      : Filler<Dtype>(param) {}
  virtual void Fill(Blob<Dtype>* blob) {
    Dtype* data = blob->mutable_cpu_data();
    CHECK(blob->count());
    caffe_rng_gaussian<Dtype>(blob->count(), Dtype(this->filler_param_.mean()), //产生高斯分布的随机数初始化
        Dtype(this->filler_param_.std()), blob->mutable_cpu_data());  //默认的高斯分布为 N~(0, 1)
    int sparse = this->filler_param_.sparse();
    CHECK_GE(sparse, -1);
    if (sparse >= 0) {  //sparse 默认为 -1
      // Sparse initialization is implemented for "weight" blobs; i.e. matrices.
      // These have num == channels == 1; width is number of inputs; height is
      // number of outputs.  The 'sparse' variable specifies the mean number
      // of non-zero input weights for a given output.
      CHECK_GE(blob->num_axes(), 1);
      const int num_outputs = blob->shape(0);
      Dtype non_zero_probability = Dtype(sparse) / Dtype(num_outputs);
      rand_vec_.reset(new SyncedMemory(blob->count() * sizeof(int)));
      int* mask = reinterpret_cast<int*>(rand_vec_->mutable_cpu_data());
      caffe_rng_bernoulli(blob->count(), non_zero_probability, mask);
      for (int i = 0; i < blob->count(); ++i) {
        data[i] *= mask[i];
      }
    }
  }

 protected:
  shared_ptr<SyncedMemory> rand_vec_;
};

/**
 * @brief Fills a Blob with values @f$ x \sim U(-a, +a) @f$ where @f$ a @f$ is
 *        set inversely proportional to number of incoming nodes, outgoing
 *        nodes, or their average.
 *
 * A Filler based on the paper [Bengio and Glorot 2010]: Understanding
 * the difficulty of training deep feedforward neuralnetworks.
 *
 * It fills the incoming matrix by randomly sampling uniform data from [-scale,
 * scale] where scale = sqrt(3 / n) where n is the fan_in, fan_out, or their
 * average, depending on the variance_norm option. You should make sure the
 * input blob has shape (num, a, b, c) where a * b * c = fan_in and num * b * c
 * = fan_out. Note that this is currently not the case for inner product layers.
 *
 * TODO(dox): make notation in above comment consistent with rest & use LaTeX.
 */
template <typename Dtype>
class XavierFiller : public Filler<Dtype> {
 public:
  explicit XavierFiller(const FillerParameter& param)
      : Filler<Dtype>(param) {}
  virtual void Fill(Blob<Dtype>* blob) {
    CHECK(blob->count());
    int fan_in = blob->count() / blob->num();   //即 fan_in = channel * height * width
    int fan_out = blob->count() / blob->channels();
    Dtype n = fan_in;  // default to fan_in 即variance_norm默认值为FAN_IN
    if (this->filler_param_.variance_norm() ==  
        FillerParameter_VarianceNorm_AVERAGE) {
      n = (fan_in + fan_out) / Dtype(2);
    } else if (this->filler_param_.variance_norm() ==
        FillerParameter_VarianceNorm_FAN_OUT) {
      n = fan_out;
    }
    Dtype scale = sqrt(Dtype(3) / n);   //scale = sqrt(3/n), n默认为 blob 的 num
    caffe_rng_uniform<Dtype>(blob->count(), -scale, scale,
        blob->mutable_cpu_data());   //产生均匀分布[-scale, scale] 的随机数初始化
    CHECK_EQ(this->filler_param_.sparse(), -1)
         << "Sparsity not supported by this Filler.";
  }
};