Are minirocket.py and minirocket_dv.py the implementation for the same Mini-Rocket algorithm?

[dbq233]

No description provided.

[angus924]

Hi @dbq233, they are very similar. Both do essentially the same thing. The key difference is that the main MiniRocket implementation (minirocket.py) samples bias values from randomly-selected training examples, whereas the deterministic variant (minirocket_dv.py) samples bias values using the entire training set. So there is some randomness in the default implementation (minirocket.py), but the deterministic variant (minirocket_dv.py) will always produce the same result. Overall, the accuracy of both approaches is almost exactly the same. The deterministic variant is slower, because it has to do a lot more work to extract the bias values.