Architecture of FCNs Flashcards
How does deep learning differ from conventional machine learning?
Deep learning automatically learns features from raw data, while conventional ML requires handcrafted feature extraction.
What is the purpose of gradient descent in linear regression?
It minimizes the least square error by updating weights iteratively.
Is K-Means clustering supervised or unsupervised learning?
Unsupervised learning.
Why is deep learning considered a form of representation learning?
It learns hierarchical representations automatically from data.
What is the visual cortex V1 responsible for?
Discriminating basic properties like orientation, spatial frequency, and color.
What are the key components of a Fully Connected Network (FCN)?
Input layer, hidden layers, and output layer.
What does backpropagation compute?
The gradient of the loss function to adjust model parameters.
What loss function is used for binary classification?
Binary cross-entropy.
What loss function is used for multi-class classification?
Categorical cross-entropy.
What is the role of SGD in training neural networks?
It updates model parameters step-by-step using the gradient of the loss function.
What is the MNIST dataset used for?
Handwritten digit recognition.
What activation function is used in the output layer for multi-class classification?
Softmax.
What does the model.fit() function do in Keras?
It trains the model using specified epochs and batch sizes.
What metric is commonly used to assess classification performance?
Accuracy
What is the role of centroids in K-Means clustering?
Centroids represent the center of each cluster and are updated iteratively to minimize intra-cluster distance.
What is representation learning?
A machine learning technique where useful features are automatically learned from raw data rather than being manually designed.
