Resolved: Problems with the final exam of the data scientist career track
Good morning. First, I'd like to say one thing: 130 minutes for an exam is enough in general, but I think that it's impossible to solve all those 42 questions in that amount of time, since there are also several practical tasks where we have to write a lot of code (I refer to the cases about Python and Machine Learning, but it's even worse with the part about SQL). If there were only theoretical questions, maybe it would be possible, but if we have to read such long questions, to think about the solution and to write all that code, it's almost impossible to answer correctly all the 42 questions in that amount of time.
Isn't there a Q&A section to ask clarifications about final exams, included the final exam of the data scientist career track? I have already reported this thing, it would be useful. Furthermore, if we find errors or bugs in the final exams, we cannot report them.
About question 17: what is the difference between answer 2 and answer 3? "To keep only the second to last element in the list" and "to remove the first zero in the list" isn't the same thing?
About question 24: the solution says that "The highest revenue generated was 2114 by company X, and the lowest revenue generated by company Y was 1980". So why my answer, 134, is wrong?
About question 39: how do you count the hyperparameters? Hyperparameters are for example the width, the depth and the learning rate. Here the depth is 2, while the width is unknown, but there is a width for every layer. I don't know if the width of input layer and output layer are hyperparameters, but I could have for example 30 neurons for the first hidden layer and 10 for the second hidden layer: are them 2 distinct hyperparameters? Or maybe here we have just the input layer and the output layer? It's not very clear.
Thank you
Hey Alessandro,
Thank you for the valuable input regarding the career track exams!
Let me address the questions you have one by one.
1. Regarding the duration of the final exams - we will soon revisit the time allowed for completing the exams.
2. Regarding your suggestion on having a section dedicated to final exams - unfortunately, that's not something we have at the moment. If you have questions regarding this topic, do open a thread here in the Q&A hub.
3. Regarding Question 17:
The answer "To keep only the second to last element in the list" refers to keeping only the element immediately preceding the last one. On the other hand, the answer "To remove the first zero in the list" refers to removing the first element in this list.
4. Regarding Question 24:
Indeed, the correct answer here is mistaken, we will correct that - thank you for pointing it out!
5. Regarding Question 39:
The question states "How many hyperparameters are there, associated with each layer, the network, and its training as a whole?". Since we are interested in each layer, we can say that the width of each layer is a separate hyperparameter. Given that there are 2 layers, that makes 2 hyperparameters so far. However, we know that the optimizer is SGD, which has a fixed learning rate. This learning rate is also considered to be a hyperparameter, thus, the answer is 3.
Thank you again for reaching out!
Kind regards,
365 Hristina
Good morning. Thank you very much for your answer.
3. I have misunderstood the meaning of the question: I thought that "To keep only the second to last element" meant "to keep from the second to the last element", while it means "To keep only the second last element" instead, that means only the second element starting from the end. In english, I would have written "the second last" instead of "the second to last" to avoid confusion, but since I'm from Italy I could be wrong with the translations.
4. So my final vote would be slightly higher than 60%, can you correct it? Luckily I have passed the exam anyway, otherwise I could not have obtained the certificate just because of that false mistake. Just for curiosity: since I'm the first student to report this, does this mean that I'm one of the first students (if not the first one, but it seems strange) to take the final exam of the data scientist career track? I thought that this exam existed for several years.
5. Doesn't the depth of the network count as a hyperparameter, too? Or maybe here we don't have hidden layers and so we don't have a depth to specify?
Thank you very much
Hey again Alessandro,
Thank you for your response!
Regarding question 3, I understand where the confusion might come from. However, the phrase "second to last" would be the correct one to use in this case.
Regarding question 4, the exams are, in fact, a relatively new addition to the platform. Unfortunately, it would not be possible to change your result at the moment. That, however, wouldn't affect your certificate in any way.
Regarding question 5:
In general, the depth of the network could be considered a hyperparameter (although, you can say that that is part of the architecture and structure of the network itself, which is not a hyperparameter). In this case, however, the depth is fixed in the question itself - it states that this network has exactly 2 dense layers. So, that is no longer a hyperparameter, we have fixed it in the network architecture.
Hope this helps!
Kind regards,
365 Hristina
Good morning. Regarding question 5, I thought that also fixed hyperparameters were hyperparameters... I don't think that this specification has ever been explained in any lesson. Anyway, thank you for your answers!
Hey again Alessandro,
I've just understood that your final exam result has been adjusted accordingly. You should be able to see the change in your profile :)
Kind regards,
365 Hristina
Thank you very much!