AI & Machine Learning Pt 2

0.0(0)
Studied by 0 people
call kaiCall Kai
learnLearn
examPractice Test
spaced repetitionSpaced Repetition
heart puzzleMatch
flashcardsFlashcards
GameKnowt Play
Card Sorting

1/97

encourage image

There's no tags or description

Looks like no tags are added yet.

98 Terms

1
New cards

In machine learning, learning a function f(x) is best described as a process that

Maps inputs to outputs learned from data

2
New cards

Which best describes function approximation in ML?

Fitting a curve or surface to data

3
New cards

In supervised learning, the true function f(x) is

Unknown but estimated from examples

4
New cards

Why is machine learning often considered function approximation rather than exact mapping?

Real data is noisy so we estimate f(x) approximately

5
New cards

Which of these performs non-linear function approximation?

Both Polynomial Regression and Decision Tree

6
New cards

Overfitting occurs when the approximated function:

Fits the training data too closely and fails to generalize

7
New cards

A model that fits the training data too closely and fails to generalize is said to be ____.

Overfitting

8
New cards

A model that is too simple and captures too little structure in the data is said to be ____.

Underfitting

9
New cards

If a model has high bias, it usually means that it

Underfits and approximates the function poorly

10
New cards

What are true about the bias-variance tradeoff.

11
New cards

Which of the following typically has the highest function approximation capacity?

Deep Neural Network

12
New cards

A function approximator is called universal if it can

Represent essentially any target function

13
New cards

The Universal Approximation Theorem applies specifically to

Neural networks with one hidden layer

14
New cards

Why might a deep network outperform a shallow one on complex function approximation?

It can model hierarchical relationships

15
New cards

Regularization helps function approximation mainly by

Penalizing large weights

16
New cards

What is the key difference between model capacity and model generalization?

Capacity is possible complexity, generalization is accuracy on new data

17
New cards

What correctly describe overfitting and generalization

18
New cards

Why is data normalization important before function approximation with neural networks?

It keeps features on comparable scales for stable training

19
New cards

A biological neuron primarily communicates using

Electric and chemical signals

20
New cards

Which part of a biological neuron corresponds to the weights in an artificial neuron?

The synapses

21
New cards

The part of a biological neuron that receives incoming signals is the ____.

Dendrites

22
New cards

The part of a biological neuron that transmits signals away from the cell body is the ____.

Axon

23
New cards

The junction where one neuron connects and passes signals to another is the ____.

Synapse

24
New cards

The biological neuron’s axon is analogous to what in an artificial neuron?

The output signal

25
New cards

What are parts of a biological neuron.

26
New cards

What are all components that make up an artificial neuron.

27
New cards

What are the correct mappings from biological neuron parts to artificial neuron parts.

28
New cards

In an artificial neuron, the weighted sum of inputs plus bias is immediately followed by

A non-linear activation function

29
New cards

In an artificial neuron, each input is multiplied by an adjustable parameter called a ____.

Weight

30
New cards

The constant term added to the weighted sum in an artificial neuron is the ____.

Bias

31
New cards

Why are activation functions like ReLU or sigmoid crucial in neural networks?

They introduce non-linearity

32
New cards

Why can’t a neural network without activation functions solve non-linear problems?

A stack of linear layers is still just a linear function

33
New cards

The main limitation of comparing artificial and biological neurons is that

Biological networks are far more complex and adaptive

34
New cards

Who first proposed the neuron model that inspired artificial neurons?

McCulloch and Pitts

35
New cards

What is true of the McCulloch-Pitts and early neuron models.

36
New cards

A single perceptron can only solve problems that are:

Linearly separable

37
New cards

The logical function that a single perceptron cannot learn because it is not linearly separable is ____.

XOR

38
New cards

The perceptron decision rule outputs class 1 when:

The weighted sum plus bias exceeds the threshold

39
New cards

Perceptron learning updates the weights only when:

The output differs from the target

40
New cards

What is true about a single perceptron.

41
New cards

The classic perceptron uses a ____ (threshold) activation that outputs 0 or 1.

Step

42
New cards

The single-neuron model with learnable weights introduced by Rosenblatt is called the ____.

Perceptron

43
New cards

The Greek letter commonly used to denote the learning rate is called

Eta

44
New cards

What does ADALINE stand for?

Adaptive Linear Neuron

45
New cards

Which concept did ADALINE contribute to modern neural networks?

Continuous weight updates and a differentiable loss

46
New cards

ADALINE adjusts its weights to minimize the mean squared ____ between its linear output and the target.

Error

47
New cards

If the learning rate is set too high in ADALINE, what typically happens?

It will fail to converge

48
New cards

When ADALINE’s weights stop changing significantly, the model has:

Converged

49
New cards

What is true about ADALINE.

50
New cards

Which algorithm generalizes both Perceptron and ADALINE ideas into multi-layer networks?

Backpropagation Neural Network

51
New cards

A neural network is mainly designed to

Approximate functions

52
New cards

The smallest computational unit in a neural network is called a

node

53
New cards

The bias term in a neuron helps by

Shifting the activation threshold

54
New cards

Which of the following architectures allows information flow from input to output without loops?

FNN

55
New cards

What is a “layer” in a neural network?

Group of neurons processing inputs

56
New cards

Which of the following is not an activation function?

MSE

57
New cards

Why do we use deep neural networks?

To learn hierarchical features

58
New cards

The capacity of a neural network increases when

number of layers increases

59
New cards

In neural networks, “epoch” refers to

One full pass of the data

60
New cards

Forward propagation refers to

Passing input data

61
New cards

Which activation function can help avoid vanishing gradients?

ReLU

62
New cards

Why does the sigmoid function sometimes cause vanishing gradients?

Gradients become small

63
New cards

Which of these can be used as an output activation for binary classification?

sigmoid

64
New cards

What is the primary role of the output layer?

Transform hidden data into final predictions

65
New cards

When performing forward propagation in batch mode

Inputs are processed in mini-batches

66
New cards

What determines the number of parameters in a layer?

Number of neurons

67
New cards

Backpropagation is used for

Updating weights

68
New cards

The loss function in training neural networks measures:

Prediction error

69
New cards

If learning rate is too high, what might occur?

Oscillations

70
New cards

What is the vanishing gradient problem?

Gradients shrink to near-zero

71
New cards

Backpropagation algorithm updates weights in which order?

From output layer to input layer

72
New cards

The derivative of the loss with respect to weights is used to

Adjust weights for minimization

73
New cards

Which of these is not required for backpropagation?

random sampling

74
New cards

Why does a neural network need nonlinear activation functions like ReLU or tanh?

They prevent stacked layers from collapsing into one overall linear mapping

75
New cards

Which expression gives the mathematical form of forward propagation for a single neuron?

output = f(Wx + b), a weighted sum plus bias passed through an activation

76
New cards

What is the key difference between softmax and sigmoid activations?

Softmax normalizes scores across classes, while sigmoid squashes each value independently

77
New cards

How does forward propagation differ between the training and inference phases?

The layer computations match, but training also stores activations for the backward pass

78
New cards

Which sequence correctly summarizes one training step with backpropagation?

Forward pass, compute loss, backward pass for gradients, then update the weights

79
New cards

Select all statements that correctly contrast single-layer and multi-layer networks.

80
New cards

Select all true statements about weights and biases in a neuron.

81
New cards

Select all statements that are true of ReLU.

82
New cards

Select all statements that are true of the sigmoid activation.

83
New cards

Select all steps that occur in one mini-batch training iteration.

84
New cards

Select all true statements about the learning rate.

85
New cards

Select all correct pairings of a task with its standard loss function.

86
New cards

Select all true statements about backpropagation.

87
New cards

The key mathematical tool used by backpropagation to compute gradients through layers is the chain ____.

rule

88
New cards

A model that fits its training data almost perfectly but performs poorly on new data suffers from ____.

overfitting

89
New cards

The activation function whose output range is 0 to 1 is the ____ function.

sigmoid

90
New cards

The activation function whose output ranges from -1 to 1 is ____.

tanh

91
New cards

For any negative input value, ReLU outputs the number ____.

0

92
New cards

The standard output activation for multiclass classification, producing class probabilities that sum to 1, is ____.

softmax

93
New cards

A fully connected layer with 4 inputs and 3 neurons has a total of ____ parameters, counting both weights and biases.

15

94
New cards

the hyperparameter that controls the step size of each gradient descent update is the learning ____.

rate

95
New cards

The optimizer that combines momentum with per-parameter adaptive learning rates is ____.

Adam

96
New cards

When gradients shrink toward zero as they flow backward through many layers, this is called the ____ gradient problem.

vanishing

97
New cards

The abbreviation for the standard regression loss, mean squared error, is ____.

MSE

98
New cards

The subset of training examples used for a single weight update in network training is called a ____.

mini-batch