Activation perform helps in capturing the non linearity within the information. Due to which Neural community understands the non linear conduct of the underlying information not like machine studying fashions. Activation perform allows the community to be taught the advanced sample within the information. There are totally different activation features obtainable:
Sigmoid:
Benefits :
- Output lies between 0 and 1.
- Sharp change within the output for the values near 0.
- Output is +1 or -1 for the values away from 0.
- Helpful for binary classification.
Disadvantages:
- Output isn’t centered at 0.
- Liable to vanishing gradient or exploding gradient downside