FINA 5840: Financial Modeling Assignment 3
(Group Homework Assignment)
Halis Sak
Due date: April 30, 2021
Question. We will work with firm characteristics data that we downloaded
from WRDS for this homework assignment.
In homework assignment 2, we observed that “mmt6”, “divyield”, and “log-
mmt” are the best features to choose if we want to use three features for linear
regression. And we learned how to use TensorFlow for implementing a two-
layer neural network in the last week’s tutorial. We want to test performance
of {“mmt6”,“divyield”,“logmmt”} under two different neural network models;
• One hidden layer of size 3 as in the case of the tutorial
• Two hidden layers of sizes 4 and 4
using again TensorFlow.
You can modify the Python notebook of the tutorial to implement these
two models. Please initialize weights of the models using np.random.uniform
function (use low=-0.01 and high=0.01). At this stage, you can print model
weights (model.weights) to understand the shape of each weight array (see In
[9] of Python notebook of the tutorial). Please report out-of-sample R2 values
for these two models.
1
学霸联盟