I'm reading the paper Pixel Recurrent Neural Network. I have a question about Row LSTM. Why Row LSTM can capture triangular contexts?
In this paper,
the kernel of the one-dimensional convolution has size $k \times 1$ where $k \geq 3$; the larger value of $k$ the broader the context that is captured.
The one-dimensional kernel can capture only the left context. (Is this correct?)
The $n \times n$ kernel such as
$$ \begin{bmatrix} 1 & 1 & 1 \\ 0 & 0 & 0 \\ 0 & 0 & 0 \end{bmatrix} $$
can capture triangular contexts.
Is this correct?
