Topic 1 Question 48
A Machine Learning Specialist is training a model to identify the make and model of vehicles in images. The Specialist wants to use transfer learning and an existing model trained on images of general objects. The Specialist collated a large custom dataset of pictures containing different vehicle makes and models. What should the Specialist do to initialize the model to re-train it with the custom data?
Initialize the model with random weights in all layers including the last fully connected layer.
Initialize the model with pre-trained weights in all layers and replace the last fully connected layer.
Initialize the model with random weights in all layers and replace the last fully connected layer.
Initialize the model with pre-trained weights in all layers including the last fully connected layer.
ユーザの投票
コメント(17)
Ans B sounds correct
👍 28rsimham2021/09/29https://www.hackerearth.com/practice/machine-learning/transfer-learning/transfer-learning-intro/tutorial/ answer is B. Read the scenario: The target dataset is large and similar to the base training dataset. Since the target dataset is large, we have more confidence that we won’t overfit if we try to fine-tune through the full network. Therefore, we:
1.Remove the last fully connected layer and replace with the layer matching the number of classes in the target dataset; 2.Randomly initialize the weights in the new fully connected layer; 3.Initialize the rest of the weights using the pre-trained weights, i.e., unfreeze the layers of the pre-trained network; 4.Retrain the entire neural network; So, first look at 3, and then at 1.
👍 3tff2021/10/11Since it is transfer learning where you retain knowledge from a solved problem, weights are to be pre-trained. So A and C are wrong. Between B and D, D keeps the last layer but that is not what you want since the question mentions a change of general objects to more specific types. So answer is B
👍 3syu31svc2021/10/16
シャッフルモード