Created by: pkuyym

ConvShift Layer is designed to implement circular convolutional operation. Circular convolutional operation is a necessary step to implement location-based addressing in Neural Turing Machine. NTM was designed to process sequence data initially. However, current ConvShift Layer only supports non-sequence type data input.
Original implementation calls circularConv and circularConvDervative functions implemented in class Matrix.

void ConvShiftLayer::forward(PassType passType) {
  Layer::forward(passType);

  MatrixPtr inV0 = getInputValue(0);
  MatrixPtr inV1 = getInputValue(1);

  size_t batchSize = inV0->getHeight();
  size_t dataDim = inV0->getWidth();

  CHECK_EQ(batchSize, inV1->getHeight());
  CHECK_EQ(dataDim, getSize());

  {
    REGISTER_TIMER_INFO("FwResetTimer", getName().c_str());
    resetOutput(batchSize, dataDim);
  }

  MatrixPtr outV = getOutputValue();

  REGISTER_TIMER_INFO("FwConvShiftTimer", getName().c_str());
  outV->circularConv(*inV0, *inV1);
}

void ConvShiftLayer::backward(const UpdateCallback& callback) {
  MatrixPtr inV0 = getInputValue(0);
  MatrixPtr inV1 = getInputValue(1);
  MatrixPtr outG = getOutputGrad();
  MatrixPtr inG0 = getInputGrad(0);
  MatrixPtr inG1 = getInputGrad(1);

  REGISTER_TIMER_INFO("BwConvShiftTimer", getName().c_str());

  if (inG0 && inG1) {
    outG->circularConvDerivative(*outG, *inV0, *inV1, *inG0, *inG1);
  } else {
    CHECK(!inG0 || !inG1) << "Not supported";
  }
}

First, I will check the type of input data. If the input data is non-sequence type, the function will run origin logic, otherwise call circularConvSeq and circularConvSeqDerivative.

void ConvShiftLayer::forward(PassType passType) {
  Layer::forward(passType);

  MatrixPtr inV0 = getInputValue(0);

  size_t batchSize = inV0->getHeight();
  size_t dataDim = inV0->getWidth();

  CHECK_EQ(dataDim, getSize());

  {
    REGISTER_TIMER_INFO("FwResetTimer", getName().c_str());
    resetOutput(batchSize, dataDim);
  }

  REGISTER_TIMER_INFO("FwConvShiftTimer", getName().c_str());
  if (!isSeqType()) {
    MatrixPtr inV1 = getInputValue(1);
    CHECK_EQ(batchSize, inV1->getHeight());
    MatrixPtr outV = getOutputValue();
    outV->circularConv(*inV0, *inV1);
  } else {
    circularConvSeq();
  }
}

void ConvShiftLayer::backward(const UpdateCallback& callback) {
  MatrixPtr inG0 = getInputGrad(0);
  MatrixPtr inG1 = getInputGrad(1);

  REGISTER_TIMER_INFO("BwConvShiftTimer", getName().c_str());

  if (!(inG0 && inG1)) {
    CHECK(!inG0 || !inG1) << "Not supported";
  }

  if (!isSeqType()) {
    MatrixPtr inV0 = getInputValue(0);
    MatrixPtr inV1 = getInputValue(1);
    MatrixPtr outG = getOutputGrad();
    outG->circularConvDerivative(*outG, *inV0, *inV1, *inG0, *inG1);
  } else {
    circularConvSeqDerivative();
  }
}

Please go to related PR #2133 to check implementation details.