Hi Everybody I am trying to build a neural network using FPGA. As you know, the one problem that should be solved is implementation of non-linear activation function. I have read thro' many papers; some implemented using look up tables, some calculated it using piecewise approximation methods. Also I have been thinking about using different number systems that would give good precision. Any suggestions or comments helping me to find more reasonable methods would be appreciated Thank you for your replies Bye Ram
- posted
20 years ago