Brainscape's Knowledge GenomeTM


Top Flashcards (Certified)
All Flashcards

ML > ML Soc-Tech > Flashcards

ML Soc-Tech Flashcards

(83 cards)

Start Studying
1
Q

What is the primary focus of statistical models?

A

Inference, such as determining whether a relationship exists or not

This involves hypothesis testing.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

What do theories describe in the context of statistical models?

A

First principles of how and why X causally relates to Y

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

What are models in the context of theories?

A

Instantiations of theories that provide a local mathematical description or understanding of a phenomenon

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

What is the significance of parameters in models?

A

Parameters are interpretable and models are tractable

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

Why is the number of parameters limited in statistical models?

A

To avoid overfitting and enhance model interpretability

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

What do the Law of Large Numbers and Central Limit Theorem indicate?

A

As sample size increases, the sample mean converges to the population mean and approaches a normal distribution

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

What is the main focus of machine learning (ML)?

A

Prediction, specifically input-output relationships

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

What is prioritized in ML over interpretability or inference?

A

Prediction accuracy on unseen data

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

What is a tractable system, model, or problem?

A

One that can be solved or analyzed using existing mathematical or computational methods within a reasonable time and resource limit

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

What does learning in machine learning involve?

A

Approximating a function g that maps input x to output y by analyzing patterns within data

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

What is the goal of learning in ML?

A

To fit the training data and generalize to unseen examples

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

What is the difference between learning and estimation?

A

Learning prioritizes developing a predictive model, while estimation focuses on determining specific parameters within a model

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

What type of data does supervised learning work with?

A

Labeled data (x, y)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

What is the purpose of classification models in supervised learning?

A

To categorize data into predefined classes

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

What is the output of regression models in supervised learning?

A

A continuous value or quantity

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

What is generalization in machine learning?

A

A measure of how effectively a model captures underlying patterns in data rather than memorizing specific details

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
17
Q

What causes poor generalization in models?

A

Overfitting or underfitting

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
18
Q

What is overfitting?

A

Occurs when a model is too complex, capturing noise in the training data instead of general patterns

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
19
Q

What is underfitting?

A

When a model is too simplistic and fails to capture significant trends in the data

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
20
Q

What does the bias-variance trade-off describe?

A

The balance between a model’s capacity to generalize and its ability to fit the training data

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
21
Q

What does high bias indicate in a model?

A

The model is too simple and fails to learn relationships in the training data effectively

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
22
Q

What does high variance indicate in a model?

A

The model is overly complex and captures noise in the training data

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
23
Q

What are the steps in ML model development?

A
  • Study phenomenon & clean data
  • Discover data
  • Explore associations
  • Train ML model
  • Evaluate model
  • Analyze errors
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
24
Q

What are the assumptions on data in regression?

A
  • Linearity
  • Constant variance (homoskedasticity)
  • Errors are independent and identically distributed
  • No correlation between errors
  • No perfect collinearity in features
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
25
What is a key advantage of decision trees?
Easy to use with minimal pre-processing required
26
What is a disadvantage of decision trees?
Sensitive to overfitting and can produce unstable results
27
What are the key components of a decision tree?
* Root node * Split node * Branch * Leaf node
28
What does entropy measure in a decision tree?
The level of impurity or randomness in the dataset
29
What does information gain indicate in decision trees?
The reduction in entropy achieved by splitting the dataset on a particular feature
30
What is pruning in decision trees?
The process of intervening to prevent overfitting
31
What are the types of pruning in decision trees?
* Pre-pruning * Post-pruning
32
What is feature importance in decision trees?
Quantifies the contribution of each feature to the predictive power of the model
33
What are common performance metrics for regression?
* Mean Squared Error (MSE) * Explained Variance * Mean Absolute Error (MAE)
34
What are common performance metrics for classification?
* Accuracy * Precision * Recall * Matthew’s Correlation Coefficient (MCC)
35
What are artificial neural networks (ANNs) inspired by?
The structure of the human brain
36
What is a major advantage of ANNs?
Highly versatile and can model non-linear relationships
37
What is a disadvantage of ANNs?
Opaque, making it unclear what the model has learned
38
What happens during training in ANNs?
Weights and biases are adjusted to identify useful patterns in the data
39
What is the role of hidden layers in ANNs?
They process data through neurons using weights, biases, and activation functions
40
What is the purpose of the output layer in ANNs?
To produce the final prediction
41
What is a loss function in ANNs?
Measures the difference between predicted and actual values
42
What do bias nodes ensure in a neural network?
Bias nodes ensure input, even if input space is 0
43
What is the purpose of the loss function in a neural network?
Measures the (global) performance of the model given the data
44
What types of loss functions are used for regression and classification?
* Regression: MSE, RMSE * Classification: Cross-entropy
45
What does UAT (Cybenko’s theorem) state about feedforward neural networks?
Can approximate any continuous function on a closed and bounded domain
46
What are the three key pre-processing steps for training ANNs?
* Features must be scaled * Categorical features must be dummy-encoded * Missing values must be handled
47
What is the purpose of backpropagation in training ANNs?
Computes how the total loss is generated from each node in each layer
48
Define the term 'epoch' in the context of training neural networks.
One full pass through the entire training dataset
49
What is the difference between 'iteration' and 'epoch'?
Iteration: One forward and backward pass through network for a batch
50
What are the three types of gradient descent?
* Batch gradient descent * Stochastic gradient descent * Mini batch gradient descent
51
What does L1 regularization penalize?
The sum of the absolute values of the weights
52
What is the purpose of hyperparameter tuning?
Finding the optimal combination of hyperparameters to improve model performance
53
What is K-fold cross validation used for?
To more rigorously evaluate the generalisation performance
54
What is an ensemble model?
A collection of sub-models (weak models) that combine to become more powerful
55
What is bagging in the context of ensemble learning?
A technique used to reduce variance and improve robustness by creating multiple versions of a dataset
56
What is the key benefit of random patching?
Reduces correlation between trees by using diverse feature subsets
57
What is boosting in ensemble learning?
An ensemble technique where models are trained sequentially to correct the errors of their predecessors
58
What is the role of embeddings in machine learning?
To represent discrete data in a continuous, lower-dimensional vector space
59
What conditions must be met to establish causality?
* Association * Temporal precedence * Nonspuriousness
60
Define 'in-distribution' prediction.
A prediction task where the test data is drawn from the same distribution as the training data
61
What is the difference between interpretability and explainability?
* Interpretability: Passive characteristic of a model * Explainability: Active characteristic of a model
62
What is the difference between interpretability and explainability in AI models?
Interpretability is a passive characteristic of a model that refers to how well a model makes sense to a human observer, while explainability is an active characteristic that involves actions taken by a model to clarify its internal functions.
63
List the properties of explanations in AI.
* Accuracy * Fidelity * Consistency * Stability * Comprehensibility * Certainty/Novelty * Degree of Importance * Representativeness
64
True or False: Fidelity refers to the discrepancy between the explainable and original model outputs.
False
65
What does simulatability refer to in the context of transparency in ML models?
The ability of a model to be simulated or comprehended entirely by a human.
66
Define False Positive Rate (FPR) in COMPAS parity testing.
The proportion of individuals who did not recidivate who were incorrectly predicted to recidivate.
67
What is the purpose of post-hoc explainability?
To identify simplified/transparent models representing specific sections of or approximating a more complex model.
68
Fill in the blank: Anchors generate ________ to explain the behavior of a machine learning model for a specific prediction.
if-then rules
69
What are counterfactual explanations?
Alternative data points that would change the model’s prediction to a desired outcome.
70
What is a Partial Dependence Plot (PDP)?
Visualizes the relationship between a feature and the predicted outcome while marginalizing over other features.
71
List the advantages of using a Partial Dependence Plot.
* Easy to understand * Simple to implement * Clear interpretation if features are uncorrelated * Provides a causal interpretation for the model
72
What are Individual Conditional Expectation (ICE) curves?
Graphs showing how predictions for individual instances change as a feature varies while keeping other features fixed.
73
What does LIME stand for?
Local Interpretable Model-agnostic Explanations.
74
True or False: LIME uses a linear approximation of the model locally around the instance being explained.
True
75
How does SHAP calculate the SHAP value?
As the change in the expected model prediction when conditioning on that feature, comparing the prediction with and without the feature.
76
What does the First-Order Sobol Index measure?
The direct contribution of an input to the variance of the output, ignoring interactions with other inputs.
77
List the ethical concerns regarding Responsible AI.
* Potential for misuse * Privacy invasion * Reinforcement of stereotypes
78
What are the risks associated with Large Language Models (LLMs)?
* Discrimination, hate speech * Information hazards * Malicious uses * Environmental and socioeconomic harms
79
What are the banned applications under the EU A.I. Act?
* Biometric categorization systems using sensitive characteristics * Untargeted scraping of facial images * Emotion recognition in workplaces and educational institutions * Social scoring based on behavior or characteristics
80
What role does XAI play in climate change?
* Forecasting electricity supply * Reducing waste in electricity grids * Predictive maintenance in transport
81
What is the significance of the SHAP Waterfall Plot?
Visualizes how each feature contributes to an individual prediction by showing the additive contributions of features step-by-step.
82
What does the SHAP Beeswarm Plot summarize?
Gives a global view of feature importance across the entire dataset by summarizing all individual SHAP values.
83
Fill in the blank: The longer the bar in the SHAP Bar Plot, the more ________ the feature is.
important

ML (1 decks)

Brainscape helps you reach your goals faster, through stronger study habits.
© 2025 Bold Learning Solutions. Terms and Conditions