Northeastern University
College of Professional Studies
ALY6020 – Predictive Analytics
Winter 2020 CPS Quarter – First Half
Assignment: Feedforward neural networks
Due date: 02/07/20
Goal
Train a feedforward neural network to implement a binary BCD (binary coded decimal) adder.
For some background, please read https://www.geeksforgeeks.org/bcd-adder-in-digital-logic/
(focus on what a BCD adder does, not on how to implement it with digital circuits. We are going
to solve the problem with a neural network.)
You will need to create a truth table that represents the mapping between all the possible inputs
and desired outputs which you want your neural network to learn.
Specific Instructions
Write a script based on mine that transforms your 8 bit input (4 bits per digit) into a 5 bit output.
As we discussed, there are no general rules that dictate how to design a network for a particular
task. Therefore you have to try different numbers of hidden layers, neurons per layer, and
activation functions (although, sigmoid functions make sense for this problem given the binary
value of the inputs and outputs). Start simple and only make your network more complicated as
necessary.
If necessary, please add QQ: 99515681 or email: [email protected] micro letter: codehelp