joelowj/Machine-Learning-and-Reinforcement-Learning-in-Finance

Bank defaults prediction using FDIC dataset (80/100)

joelowj opened this issue · 2 comments

joelowj commented

Current submission

Logistic Regression Stats Model 20/20
Logistic Regression using Scikit-learn 20/20
Logistic Regression fewer predictors 20/20
Logistic Regression with Tensorflow 0/20
Logistic Regression with Neural Network 20/20

I am left with Logistic Regression with Tensorflow which did not clear the grader (your answer to Logistic Regression with TensorFlow is not correct). In case, there is any issue with the grader, I will wait for a while before attempting this question again. If anyone managed to clear this and would like to help, feel free to start a discussion.

randy-tsukemen commented

Hi, I found that the grader need the gradient to run n_epochs*n_batches times. But I think the random_batch function did not go through all X_trainTf is rarely seen in other ML implementation.

for epoch in range(n_epochs):        
        for batch_index in range(n_batches):
            X_trainTf_batch, y_trainTf_batch = random_batch(X_trainTf, y_trainTf, batch_size)
            sess.run([optimizer, loss], feed_dict={X: X_trainTf_batch, 
                                                   y: y_trainTf_batch})
    y_proba_val = sess.run(y_proba, feed_dict={X: X_testTf})
joelowj commented

@minori111 , I see. Would you be able to post a screen shot of this set of code clearing the grader? And if so, would you like to make a pull request for it? (: