Deep Learning Flashcards
In the backpropagation process, what does the term “gradient” refer to?
The slope of the loss function at a certain point
In the context of backpropagation, what happens when a neural network is said to be “converging”?
The network is learning and the error is decreasing over iterations
What is a common problem that can occur during backpropagation, especially with deep neural networks?
Gradient vanishing/exploding
‘Underfitting’ in a machine learning model is a situation where:
The model has not learned the training data well enough
What role does backpropagation play in addressing overfitting or underfitting in a neural network?
It helps to calculate gradients for updating weights, contributing to the learning process
In unsupervised learning, what does the model learn from?
From only unlabeled data
Which of the following is a common application of unsupervised learning?
Customer segmentation
Which of the following is an example of a problem best suited for reinforcement learning?
Playing a game of chess
Which learning method often involves the use of a ‘reward function’?
Reinforcement learning
What is the role of a ‘label’ in supervised learning?
It is the desired output value for a given input data point
What differentiates deep learning from traditional machine learning?
Deep learning algorithms are based on the structure and function of the human brain
Why is the term ‘deep’ used in ‘deep learning’?
Because it refers to algorithms that use a deep stack of transformations from input to output
What is the primary difference between traditional machine learning and deep learning?
Deep learning algorithms require more data
How does the Transformer model, used in natural language processing tasks, handle sequential data?
It processes the sequence all at once using self-attention mechanisms
Which of the following tasks can deep learning assist with in the field of natural language processing?
A. Spelling correction
B. Sentence generation
C. Machine translation
D. All of the above
What is the role of word embeddings in natural language processing?
They are used to convert words into numerical vectors that capture semantic meaning
Consider a scenario where you are analyzing tweets to determine overall sentiment (positive, negative, or neutral). Which type of sequence-to-sequence model configuration would be most appropriate for this task?
Many-to-One
What is the main concept behind Generative Adversarial Networks (GANs)?
Two models, a generator and a discriminator, are trained together where the generator tries to fool the discriminator and the discriminator tries to not get fooled
What is the Transformer model primarily used for?
Natural Language Processing (NLP)
Which part of a GAN is responsible for generating new data?
The generator
Which part of a GAN is trained to distinguish real data from fake data?
The discriminator
What is one major advantage of Transformers over Recurrent Neural Networks (RNNs) for sequence prediction tasks?
Transformers can process longer sequences without forgetting earlier parts of the sequence
What part of a neuron is similar to the activation function in a neural network?
Soma
What type of artificial neural network architecture was inspired by the visual cortex?
Convolutional Neural Network (CNN)
