Open des-zhong opened 5 months ago
It is hard to handle regression tasks in ANN2SNN. One explaination is that regression results are sensitive with conversion error because they are specific values but for classification tasks, the specific values are not that important.
I have trained a ANN of my own for a regression problem and I'm thinking about transforming the trained ANN to a SNN using the "ann2snn.Converter" method in spikingjelly. But the results seems incorrect. The predicted vector of SNN is order-preserve but the absolute value is far from correct. How should I transform a regression ANN to SNN?