Target data in neural network

I have just started neural network and deeplerning course. I have completed the assignments to create and train the model for image classification. That worked successfully.

I dont understand one thing. Input dataset is images of cat, which will be converted to vectors as input of NN. After all computation I will have predicted output. According to logic predicted output will be matched with target data. I find that the labels are target data. But labels are string (e.g cat), but how this string data can be matched with predicted output (vector) . What is this target data? need to learn in details.

We also convert labels to numbers, for example, 0 for cat and 1 for dog. So, when the model output is 1, we know it means a dog.

PS: You posted this in the General Discussions category. Please use the pencil icon along the title to move this thread to the proper course category.

Right! And once we have converted the labels to numeric values, we have the “cross entropy” loss functions for either the binary case (“yes/no” or “cat/not cat”) and the multiclass case (cat, dog, horse, cow, elephant, kangaroo …). The loss function gives you a numeric output that expresses the distance between the current output of your model and the desired (correct) answer for each training sample.

I think we ought to switch to examples that use elephants and kangaroos, instead of cat / not_cat.