neural-networks-deep-learning

AI Essentials - Grok · 30 questions

• In a multilayer neural network, why are nonlinear activation functions applied after (or as part of) each neuron's weighted sum?
• During training of a very deep network, gradients for the earliest layers become vanishingly small so those weights barely update. What problem is being described?
• In neural network training, what is the primary purpose of the backpropagation algorithm?
• Compared with smaller models, what is a well-known disadvantage of training a very large, deep neural network?
• What is a defining characteristic of a graph neural network (GNN)?
• Which description best captures transfer learning in modern machine learning practice?
• For which class of problems are recurrent neural networks (RNNs) primarily designed?
• During neural network training, which statement best describes the primary job of the loss function?
• Why do recurrent neural networks use backpropagation through time (BPTT) when training?
• In a standard multilayer neural network, what is the role of a hidden layer?
• Which description correctly defines a feedforward neural network?
• Which statement best describes a defining trait of a Long Short-Term Memory (LSTM) network?
• A data scientist is building a model that outputs the probability a customer will click an advertisement. Which activation is appropriate for the final layer?
• A studio wants a model that predicts the next frame in a video given preceding frames. Which architecture family is the most appropriate starting point?
• Why do deep networks rely on nonlinear activation functions between layers rather than purely linear transforms?
• In deep learning frameworks, what does the term tensor most accurately refer to?
• During neural network training, which procedure computes how much each weight contributed to the loss by propagating error signals from the final layer back through earlier layers?
• In a multiclass neural classifier, what is the usual role of the softmax function in the final layer?
• Which limitation is most characteristic of a plain, vanilla RNN when modeling lengthy sequences?
• How does a convolutional layer fundamentally differ from a fully connected (dense) layer in a neural network?
• What is the main purpose of the sigmoid activation function in neural networks?
• Which statement best describes an artificial neural network (ANN)?
• What is the vanishing gradient problem, and in which setting is it especially severe?
• How does the ReLU (Rectified Linear Unit) activation behave?
• In multi-class classification, what does a final softmax layer produce?
• When building a sequence model for language tasks, why might a practitioner prefer a Gated Recurrent Unit (GRU) architecture over a plain recurrent neural network (RNN)?
• How does a long short-term memory (LSTM) network primarily differ from a simple recurrent neural network when modeling long sequences?
• In a convolutional neural network, what is the fundamental structural difference between a fully connected layer and a convolutional layer?
• What does the exploding gradient problem describe in deep network training?
• What defines a fully connected neural network (dense multilayer network) at the connectivity level?