Closed naveedunjum closed 7 months ago
I think there are a lot of ways to approach this. I recommend you look at the sigma delta tutorial here which predicts a continuous steering angle.
This is not the only way to get a continuous like value. For example, you can calculate spike rates over a sliding window. Then you can transform or interprete that rate however you want.
For example, If you introduce a bias term, you can constrain the output of the SNN to produce values between 0 and 1 with a sigmoid as shown in this demo.
I think there are a lot of ways to approach this. I recommend you look at the sigma delta tutorial here which predicts a continuous steering angle.
This is not the only way to get a continuous like value. For example, you can calculate spike rates over a sliding window. Then you can transform or interprete that rate however you want.
For example, If you introduce a bias term, you can constrain the output of the SNN to produce values between 0 and 1 with a sigmoid as shown in this demo.
Thanks. I will have a look at this. I will post here how I far I can go with this
User story
As a user, I want to request some ways to solve regression problems using Slayer. I tried the xor_regression tutorial given and modifying that, by changing the dataloader for the dataset. But my outputs were always zero, and the network wasnt learning anything.
For example I wanted to recreate a sinusoidal function using Slayer, so I used a single dimension x axis to predict the corresponding sin value. But the outputs were always zero.
Can someone please help with dealing with outputs with continous functions?