ExamPark
Course menuAIF-C01

Exams / AIF-C01 / Study guide / AI, ML and deep learning basics

AI, ML and deep learning basics: AIF-C01 study notes and practice questions

What to know about AI, ML and deep learning basics for the AIF-C01 exam, from the Fundamentals of AI and ML domain (20% of the exam guide), then 3 free questions to test it. Part of the AIF-C01 study guide.

Last verified against current AWS documentation ·How we write our questions

What to know for the exam

This topic is the vocabulary everything else builds on. Questions give a description and ask for the term, or describe data and ask what kind it is.

How the terms fit together

  • Artificial intelligence (AI): machines doing tasks that normally need human intelligence.
  • Machine learning (ML): a way to build AI by learning patterns from data instead of writing rules by hand.
  • Deep learning: ML with neural networks of many layers; it powers image recognition, speech and language models.
  • Generative AI: deep learning models that create new content (text, images, code) rather than only classifying or predicting.

Key terms

Term Meaning
Model What training produces: a function that turns inputs into predictions
Algorithm The method used to train a model
Training Feeding data to an algorithm so the model learns patterns
Inference Using a trained model to make predictions on new data; can be real-time or batch
Feature An input variable, such as age or price
Label The known answer in training data
Fit How well a model matches the data: overfitting learns training data too closely, underfitting misses the pattern
Bias A systematic error, from skewed data or assumptions
Large language model (LLM) A very large model trained on text to understand and generate language

Types of learning

Type Data Example
Supervised Labeled examples Predict whether a transaction is fraud from past labeled transactions
Unsupervised No labels Group customers by similar behavior (clustering)
Reinforcement Rewards for actions Train a robot or game agent by trial and error

Types of data

Labeled vs unlabeled; structured (tables) vs unstructured (text, images, audio); tabular, time-series (ordered measurements over time), image and text data.

Common traps

  • Deep learning vs ML. All deep learning is ML, but not all ML is deep learning.
  • Supervised vs unsupervised. If the data has the answer you want to predict, it’s supervised.
  • Training vs inference. Inference is using the model, not building it.

Try the 3 practice questions

Practice AI, ML and deep learning basicsTake the free practice exam

Question 1

Easier ·Link to this question (question 1)

A retail company's planning team keeps mixing up AI terms in strategy meetings. An analyst explains that a fraud model uses many stacked layers of artificial neurons to learn patterns from millions of transactions.

Which statement correctly places deep learning among the other terms?

Choose one answer.

  1. A. Deep learning is a broader field than AI, and AI is one of its narrower techniques for finding patterns.
  2. B. Deep learning is a subset of machine learning that uses neural networks with many layers.
  3. C. Deep learning is a rule-based approach that needs no training data and is written as fixed if-then logic.
  4. D. Deep learning is another name for generative AI, used only to write text, images and code for users.
Show the answer and explanation

Answer: B

  • A. Wrong: Reversed. AI is the umbrella term, machine learning sits inside it, and deep learning sits inside machine learning.
  • B. Right: Correct. Deep learning is a part of machine learning built on artificial neural networks with many layers of neurons. Those layers learn patterns directly from large amounts of data. Learn more: What is deep learning? (opens in a new tab)
  • C. Wrong: Rule-based systems follow hand-written logic. Deep learning learns its behavior from training data, so it is not a rule-based approach that works without data.
  • D. Wrong: Generative AI is built on deep learning, but deep learning also powers image recognition, speech recognition and forecasting. It is not limited to one use.

Explanation

AI is the broad goal of machines performing tasks that normally need human intelligence. Machine learning is one way to reach it, by learning patterns from data. Deep learning is a subset of machine learning that uses many-layered neural networks, and generative AI is built on top of it. This question tests how the terms nest inside one another.

Study

Question 2

Easier ·Link to this question (question 2)

A gym chain holds membership and visit records for 200,000 customers, but nobody has labeled them with customer types.

Marketing wants the customers split into groups of similar people so it can design a campaign for each group. It has no outcome to predict.

Which machine learning approach fits?

Choose one answer.

  1. A. Supervised classification trained on labeled customer types.
  2. B. Reinforcement learning that rewards the model for each new sign-up.
  3. C. Regression that predicts each customer's monthly spend.
  4. D. Unsupervised clustering that groups customers by similarity.
Show the answer and explanation

Answer: D

  • A. Wrong: Classification needs labeled examples of each type. The gym has no labels, so the model cannot be trained this way.
  • B. Wrong: Reinforcement learning learns by acting and receiving rewards. Counting sign-ups does not split existing customers into segments.
  • C. Wrong: Regression predicts a number for each customer. It does not divide the customers into groups of similar people for campaigns.
  • D. Right: Correct. Clustering finds groups of similar records without labels, which is what the gym needs. Each group can then receive its own campaign. Learn more: K-Means Algorithm in Amazon SageMaker AI (opens in a new tab)

Explanation

Clustering is unsupervised learning, so it works on data with no labels and no target to predict. It groups records by how similar they are. This question tests choosing clustering when the business needs segments rather than a prediction.

Study

Question 3

Easier ·Link to this question (question 3)

A water utility records the pressure in every pipe once a minute and wants to forecast pressure for the next hour to prevent bursts.

Each reading is stored with its timestamp, and the order of the readings matters.

How is this data BEST described?

Choose one answer.

  1. A. Image data, because the readings can be drawn as a line chart for the operators.
  2. B. Unstructured text data, because the readings are logged as messages from each pipe.
  3. C. Categorical data, because each pipe falls into a fixed group, such as its material.
  4. D. Time-series data, because the values are recorded and ordered by time.
Show the answer and explanation

Answer: D

  • A. Wrong: Image data is made of pixels from photos or scans. A chart drawn from the readings does not turn the readings into an image.
  • B. Wrong: Unstructured text is free-form writing such as emails or reports. Numeric sensor readings are structured, even when they are logged.
  • C. Wrong: Categorical data places items into named classes, such as pipe material. Pressure values are continuous numbers recorded over time.
  • D. Right: Correct. Time-series data is a sequence of values indexed by time, and forecasting relies on the order of past readings. Learn more: What is Amazon Forecast? (opens in a new tab)

Explanation

Time-series data is a sequence of measurements recorded at regular or irregular points in time, and the ordering carries the information a forecast needs. The question tests recognizing the data type from how it is recorded and what it will be used for.

Study

Want more on AI, ML and deep learning basics? The 5 full-length practice sets have 15 more AI, ML and deep learning basics questions, 325 in all, none of them in the free exam. One payment of $14.99, no subscription. See the practice sets

Study AI, ML and deep learning basics

More Fundamentals of AI and ML topics