Can someone please help me with creating my own models i mean i have been learning and doing assignments but creating models with my own data is very hard can someone please help ??
I will help : make sure that problem statement and how to solve it which model is better what are requirement topics for that model like regression we have cost function ,gradient desent …but the dataset must be preprocessed
Oh thank you a lot Bro what I struggle mostly with is the shapes like I mismatch them …
Have you gotten to the TensorFlow material in DLS C2 W3 yet? The best idea for creating your own solutions is to use TF, rather than trying to code everything yourself in python. Also note that DLS C2 W3 is just a very short intro to TF. There is a lot more to learn and more techniques are covered in DLS C4. You can also look at the TensorFlow Specializations to get more experience with how to apply TF to various kinds of problems.
Oh thank you sir because I just realised that coding everything is just a lot of work and memorizing so I wanna try that tensor flow specialization so that I can create models with it but I wanna finish that course first with the best understanding of neural networks so that I might jump to tensor flow specialization when I’m done and I’ll also just try creating easymodels in tensorflow for now…
Yes, that sounds like a great plan to finish at least DLS C2 and then look at the TF courses. You might also want to take DLS C4 first, since ConvNets are really important for lots of image recognition and computer vision applications.
But it definitely is true that even professionals doing DL work in real applications use one of the frameworks like TF or PyTorch or one of the others for building real solutions. The State of the Art these days is so deep that there’s no way any one person can write all the code you really need by hand. Prof Ng does have a good pedagogical reason for showing us how to code the core algorithms in python first, because that gives us much better intuition for what is going on. That will help a lot when it comes time to put together real solutions. If you only treat the components as “black boxes” and don’t understand what they are doing inside, it really hampers your abilities. The pattern in DLS C4 will be exactly the same as it was in C1 and C2: first he’ll walk us through how the core ConvNet algorithms work and have us build those in python and then he switches to using TF when it comes time to put together complex solutions.
Okay thank you a lot sir this is the type of help I needed may God be with you.