Trait collenchyma_nn::Pooling

pub trait Pooling<F>: NN<F> {
    fn new_pooling_config(&self, window: &[i32], padding: &[i32], stride: &[i32]) -> Result<Self::CPOOL, Error>;
    fn pooling_max(&self, x: &mut SharedTensor<F>, result: &mut SharedTensor<F>, config: &Self::CPOOL) -> Result<(), Error>;
    fn pooling_max_plain(&self, x: &SharedTensor<F>, result: &mut SharedTensor<F>, config: &Self::CPOOL) -> Result<(), Error>;
    fn pooling_max_grad(&self, x: &mut SharedTensor<F>, x_diff: &mut SharedTensor<F>, result: &mut SharedTensor<F>, result_diff: &mut SharedTensor<F>, config: &Self::CPOOL) -> Result<(), Error>;
    fn pooling_max_grad_plain(&self, x: &SharedTensor<F>, x_diff: &SharedTensor<F>, result: &SharedTensor<F>, result_diff: &mut SharedTensor<F>, config: &Self::CPOOL) -> Result<(), Error>;
}

Provides the functionality for a Backend to support Pooling operations.

Required Methods

fn new_pooling_config(&self, window: &[i32], padding: &[i32], stride: &[i32]) -> Result<Self::CPOOL, Error>

Creates a new PoolingConfig, which needs to be passed to further pooling Operations.

fn pooling_max(&self, x: &mut SharedTensor<F>, result: &mut SharedTensor<F>, config: &Self::CPOOL) -> Result<(), Error>

Computes non-linear down-sampling (max Pooling) over the input Tensor x with complete memory management.

Saves the result to result.

For a no-memory managed version see pooling_max_plain.

fn pooling_max_plain(&self, x: &SharedTensor<F>, result: &mut SharedTensor<F>, config: &Self::CPOOL) -> Result<(), Error>

Computes the max pooling over the input Tensor x without any memory management.

Saves the result to result.

Attention:
For a correct computation result, you need to manage the memory allocation and synchronization yourself.
For a memory managed version see pooling_max.

fn pooling_max_grad(&self, x: &mut SharedTensor<F>, x_diff: &mut SharedTensor<F>, result: &mut SharedTensor<F>, result_diff: &mut SharedTensor<F>, config: &Self::CPOOL) -> Result<(), Error>

Computes the gradient of max Pooling over the input Tensor x with complete memory management.

Saves the result to result_diff.

For a no-memory managed version see pooling_max_grad_plain.

fn pooling_max_grad_plain(&self, x: &SharedTensor<F>, x_diff: &SharedTensor<F>, result: &SharedTensor<F>, result_diff: &mut SharedTensor<F>, config: &Self::CPOOL) -> Result<(), Error>

Computes the gradient of max pooling over the input Tensor x without any memory management.

Saves the result to result_diff.

Attention:
For a correct computation result, you need to manage the memory allocation and synchronization yourself.
For a memory managed version see pooling_max_grad.

Implementors