I’m just starting out the course and I’m stuck at getting my system ready, and a web server is needed. The course recommends Web Server for Chrome App.
Hi I had been training the model to achieve 0.9500 accuracy and submitted the Programming Assignment: Week 1 - Breast Cancer Classification. I reached Accuracy: 0.9582 but my submission came back with the following message:
Almost. Your model has NOT passed the required level of accuracy on the test set. The minimum required accuracy is 0.9500. Please try again.
My question is, when the expected minimum acccuracy is 0.9500, the system should have accepted ‘0.9582’ as it is above minimum required. So should my model reach exact number of 0.9500? Somone kindly help me with this.
Thank you in advance.
what does your submission grader output states on why it failed ? loss of 0.48 is the concern!!! DID YOU TWEEK THE UNITS AS YOU HAD SHARED A SIMILAR THREAD ISSUE ? INCREASED OR REDUCED?? TO GET MORE LOWER LOSS.
Hi, I tweaked units many times and in between i submitted 14 times so far and i get the following from the automated grader everytime: ‘Almost. Your model has NOT passed the required level of accuracy on the test set. The minimum required accuracy is 0.9500. Please try again.’ In some cases, my loss was also very low but the accuracy exceeded and went above the mimimum expected accuracy ‘0.9500’ and sometimes it reached accuracy of 1 which i thought is overfitting, so again i tweaked learning rate as well. I have Due Jul 7, 11:59 PM CDT to complete this assignment. I will keep trying. Thank you.
also another hint for this issue would be to refer the ungraded labs and the videos where Lawrence explains you the similar codes.
probably in your code it is not only units need to be looked upon but also other parameters such as loss, optimiser or other model significant features.
I did notice a correction required in your codes which you can correct it by referring to videos again, go line by line when Lawrence is explaining you model layers as well as training and test set, you will get the solution.
If this doesn’t finds you solution, let me know. I know you might not like my response but this one assignment will teach you many important points on model creation.
You actually have idea about where you are going wrong but not getting it.
Hi DP, I will go back to Lawrences videos again. I agree, i might be missing something. Here is what i kept trying:
I increased number of units in the first hidden layer from 64 and 32 in the second layer to 128 and 64 units.
I added 3rd hidden layerwith 64 units and ReLu activation.
Increased the epochs from 200 to 1000 to gain more accuracy.
decreased the learning rate from 0.001 to 0.0005
Added L2 Regularizer to prevent overfitting.
I am happy that i am practically learning all these but just concerned if i miss the due date to submit this assignmet. Probably Lawrence videos hlep.
Thank you
Vamsi.
DP, just curious about the grading system. Is it looking for the code with the orginal comments and specifically some comments like ’ // YOUR CODE HERE’ to grade the assignment apart from the outputs?. The reason I am asking is i removed some of the comments so i can see all code in 1.5 pages. Please let me know.
Thank you
Vk
DP, just curious about the grading system. Is it looking for the code with the orginal comments and specifically some comments like ’ // YOUR CODE HERE’ to grade the assignment apart from the outputs?. The reason I am asking is i removed some of the comments so i can see all code in 1.5 pages. Please let me know.
Thank you
Vk
Sorry vamsi, am not sure if the statement you mention is between starting and ending of the codes, you are not suppose to remove, but from what I remember for these assignment I think you can remove.
Moderator DP, Finally Passed. ‘Congratulations! Your model has passed the required level of accuracy on the test set.’. My final training numbers are loss: 0.024…, acc: 0.99…
Lessons Learned: As Moderator suggested, watch the video closely and use simple approaches first and keep experimenting. Comments can be deleted but they work as guide in initial stages.
Thank you
Vk