TCN: A Causal 1D Convolution
Intermediate
The problem, from first principles
A causal 1D convolution computes each output from the current input and the few before it, never the future. It is the basic operation of a temporal convolutional network.
From theory to code
Implement causal_conv1d(x, w), the causal convolution with zeros before the start of the sequence.
Constraints
w[0]multiplies the current input.
Hints
Hint 1
For each time step, sum the kernel taps times the inputs that many steps back, skipping positions before the start.
Keywords
research-paperssequence-modelsconvolutionsequence
Python
Run your code to see results
No results yet
Run your code to check it against the examples, or submit to run every test.