Pass Oracle OCI AI Foundations — the First Time

1. A retail analytics team stores customer records in a relational database with clearly defined columns such as customer ID, purchase amount, and transaction date. Which category of data best describes this dataset?

   • A. Unstructured data
   • B. Streaming-only data
   • C. Semi-structured data
   • D. Structured data

   Answer: D. Data organized into fixed rows and columns within a relational database, with a predefined schema, is structured data. Unstructured data (images, free-text, audio) has no predefined model, and semi-structured data (JSON, XML) carries tags but no rigid table schema. 'Streaming-only' describes how data arrives, not its structure.

   Exam domain: Intro to AI Foundations

2. Which statement correctly describes the relationship among Artificial Intelligence (AI), Machine Learning (ML), and Deep Learning (DL)?

   • A. Deep Learning is a broad field that contains Machine Learning, which in turn contains AI
   • B. AI, ML, and DL are three completely separate and unrelated fields
   • C. Machine Learning is a subset of AI, and Deep Learning is a subset of Machine Learning
   • D. Deep Learning and AI are the same thing, while Machine Learning is unrelated

   Answer: C. AI is the broadest field; ML is a subset of AI that learns patterns from data; DL is a subset of ML that uses multi-layered neural networks. Option A reverses the hierarchy, and the other options misstate the nested relationship.

   Exam domain: Intro to AI Foundations

3. A bank wants to predict the exact dollar amount a customer is likely to spend next month based on historical spending. Which type of machine learning task does this represent?

   • A. Classification
   • B. Dimensionality reduction
   • C. Clustering
   • D. Regression

   Answer: D. Predicting a continuous numeric value (a dollar amount) is a regression task. Classification predicts discrete categories, clustering groups unlabeled data, and dimensionality reduction compresses features rather than predicting a target value.

   Exam domain: Intro to AI Foundations

4. A data science team has a dataset of 50,000 historical loan applications, each tagged with whether the borrower ultimately defaulted or repaid. They want to train a model that predicts default for new applicants. Which category of machine learning does this scenario describe?

   • A. Reinforcement learning, because the model learns from reward signals
   • B. Unsupervised learning, because the model discovers structure on its own
   • C. Semi-structured learning, because the data has both numbers and text
   • D. Supervised learning, because the training data includes known output labels

   Answer: D. Supervised learning trains on data where each example carries a known label (here, defaulted vs. repaid), and the model learns to map inputs to those labels. Unsupervised learning has no labels, and reinforcement learning learns from rewards through trial-and-error interaction. 'Semi-structured learning' is not a real ML category.

   Exam domain: Intro to ML Foundations

5. An autonomous warehouse robot improves its navigation policy over time by receiving positive feedback when it reaches a target shelf quickly and negative feedback when it collides with obstacles. Which type of machine learning is the robot using?

   • A. Transfer learning
   • B. Supervised learning
   • C. Reinforcement learning
   • D. Unsupervised learning

   Answer: C. Reinforcement learning is defined by an agent that learns a policy through trial and error using reward and penalty signals from its environment, exactly as described. Supervised learning requires labeled examples, and unsupervised learning finds patterns in unlabeled data. Transfer learning is a technique for reusing a pretrained model, not a feedback-driven paradigm.

   Exam domain: Intro to ML Foundations

6. A retailer has millions of customer transaction records but NO predefined customer categories. They want to automatically discover natural groupings of customers based on purchasing behavior so the marketing team can design targeted campaigns. Which learning approach is most appropriate?

   • A. Supervised learning with a classification model
   • B. Supervised learning with a regression model
   • C. Unsupervised learning with a clustering model
   • D. Reinforcement learning with a reward function

   Answer: C. Because there are no predefined labels and the goal is to discover natural groupings, this is unsupervised learning, and clustering is the technique that groups similar data points together. Classification and regression are supervised methods that require labeled training data. Reinforcement learning addresses sequential decision-making with rewards, not grouping of static records.

   Exam domain: Intro to ML Foundations

7. In a feedforward artificial neural network, what is the primary purpose of an activation function applied at each neuron?

   • A. To introduce non-linearity for modelling complex relations
   • B. To permanently store the training data inside the neuron
   • C. To convert all inputs into a fixed-size image format
   • D. To compress the model so it requires no GPU for training

   Answer: A. Activation functions (such as ReLU, sigmoid, or tanh) introduce non-linearity, allowing a network of stacked layers to approximate complex, non-linear functions; without them, any number of layers would collapse into a single linear transformation. They do not store data, reformat inputs into images, or eliminate hardware needs.

   Exam domain: Intro to DL Foundations

8. A retail team wants to automatically classify uploaded product photos by category. Which type of neural network architecture is specifically designed to excel at this kind of image-based task?

   • A. A recurrent neural network (RNN)
   • B. A k-means clustering model
   • C. A linear regression model
   • D. A convolutional neural network (CNN)

   Answer: D. CNNs use convolutional layers to detect spatial features (edges, textures, shapes) in grid-like data such as images, making them the standard choice for image classification. RNNs are designed for sequential data, linear regression predicts continuous values, and k-means is an unsupervised clustering algorithm; none are purpose-built for image recognition.

   Exam domain: Intro to DL Foundations

9. Within a CNN, what is the main role of a pooling layer (for example, max pooling) placed after a convolutional layer?

   • A. It assigns class probabilities as the final output of the network
   • B. It converts the image into a textual caption
   • C. It reduces the feature maps' spatial dimensions, cutting compute
   • D. It increases the resolution of the input image to capture more detail

   Answer: C. Pooling downsamples feature maps, shrinking their width and height to reduce the number of parameters and computation while making detected features more robust to small shifts in position. It does not upscale images, generate captions, or produce final class probabilities (that is typically the job of a softmax output layer).

   Exam domain: Intro to DL Foundations

10. A large language model (LLM) produces text one token at a time, where each new token is predicted based on all the tokens that came before it. What is this generation behavior called?

    • A. Autoregressive generation
    • B. Recurrent backpropagation
    • C. Convolutional decoding
    • D. Reinforcement sampling

    Answer: A. LLMs based on decoder transformers generate text autoregressively, predicting the next token conditioned on the preceding sequence. 'Convolutional decoding' refers to CNN-style operations not used for token generation; 'reinforcement sampling' and 'recurrent backpropagation' are not standard terms for this process.

    Exam domain: Intro to Generative AI & LLMs

Want more? Finish the quiz above to unlock 10 more free questions, or open the full 1Z0-1122-26 practice app.