Advanced Learning Algorithms Flashcards
What is a neural network?
How is a neural network structured?
What is forward propagation?
What is backward propagation?
What are activation functions, and why are they important in neural networks?
What are some commonly used activation functions?
How do you choose the right activation function for a given problem?
When should you use the sigmoid function?
When should you use the ReLU activation function?
What is the difference between linear regression and a neural network with linear activation functions?
What is the difference between binary classification and multiclass classification?
What is the softmax regression algorithm, and how is it used in multiclass classification?
What is the cross-entropy loss function, and how is it used in the context of softmax regression?
What is the indicator function used in the cost equation for Softmax regression?
How can numerical instability issues be mitigated when using the softmax activation function?
What are logits in the context of machine learning, and how are they used in the SparseCategoricalCrossentropy loss function?
What is multi-label classification, and how does it differ from single-label classification?
How can a neural network be trained for multi-label classification?
What is the Adam algorithm, and how is it used in optimization for machine learning?
How can you evaluate a models performance in machine learning?
How can you diagnose whether an algorithm has high bias or high variance?
What is regularization in machine learning?
How does L2 regularization prevent overfitting in a model?
How can regularization control both high bias and high variance in a model?
Why is it useful to compare an algorithms performance to that of a human or competing algorithms?
How can learning curves help diagnose whether an algorithm is suffering from high bias or high variance?
How can you debug a learning algorithm that is suffering from high bias or high variance?
What are precision and recall, what are they used for?
When is accuracy not the most appropriate metric to use when evaluating the performance of a machine learning algorithm?
Define the F1 score
What is a precision recall curve, and what is the significance of the area under the curve?
How are precision and recall calculated?
What is the difference between true positives and false positives?
How do precision and recall provide more insight into the performance of a model on imbalanced datasets?
What are some limitations or weaknesses of using precision and recall as error metrics?
