1z0-1110-25 Questions and Answers

Detailed Answer in Step-by-Step Solution:

Objective: Assess if bias is a common issue in data science.

Define Bias: Systematic errors in data/models (e.g., skewed training data).

Evaluate Statement:

Bias arises from unrepresentative data, poor feature selection, or algorithmic flaws—widely recognized in ML.

Examples: Gender bias in hiring models, racial bias in facial recognition.

Reasoning: Literature and practice (e.g., fairness in AI) confirm bias as prevalent.

Conclusion: A (True) is correct.

OCI documentation notes: “Bias is a common challenge in data science, stemming from imbalanced datasets or flawed assumptions, requiring techniques like re-weighting or fairness checks.” This aligns with industry standards—bias is a well-documented issue, making A true.

Oracle Cloud Infrastructure Data Science Documentation, "Addressing Bias in Models".

Detailed Answer in Step-by-Step Solution:

Objective: Scale up a notebook session without losing work.

Understand Persistence: Block volume stores session data (e.g., /home/datascience).

Evaluate Options:

A: Recreating work—inefficient, risks loss.

B: Local download/upload—cumbersome, unnecessary.

C: Use block volume persistence, scale up—efficient, preserves work—correct.

D: Object Storage—extra steps, not needed with block volume.

Reasoning: C leverages OCI’s built-in persistence for seamless scaling.

Conclusion: C is correct.

OCI documentation states: “Files in /home/datascience are stored on the block volume. To scale up, deactivate the session, provision a new one with a larger shape, and the block volume persists your work.” A loses data, B and D add complexity—only C is optimal.

Oracle Cloud Infrastructure Data Science Documentation, "Scaling Notebook Sessions".

Detailed Answer in Step-by-Step Solution:

Objective: Find the OCI service for free geospatial images.

Evaluate Options:

A: Big Data Service—Spark processing, not datasets.

B: Analytics Cloud—Visualization, not data source.

C: Data Science—ML platform, not dataset provider.

D: Open Data—Free public datasets, including geospatial—correct.

Reasoning: Open Data is OCI’s public dataset hub.

Conclusion: D is correct.

OCI documentation states: “Oracle Open Data provides free access to curated datasets, including geospatial images, for public use.” A, B, and C serve other purposes—only D delivers free geospatial data.

Oracle Cloud Infrastructure Open Data Documentation, "Dataset Offerings".

Detailed Answer in Step-by-Step Solution:

Objective: Publish a conda env to Object Storage.

Process: Initialize bucket config, then publish env.

Evaluate Options:

A: Publishes env with slug—correct final step.

B: Lists envs—unrelated to publishing.

C: Sets bucket details—required setup—correct.

D: Creates env—not publishing.

E: Activates env—not for sharing.

Reasoning: C sets up, A executes—standard workflow.

Conclusion: A and C are correct.

OCI documentation states: “To publish a conda environment, first run odsc conda init (C) with bucket namespace and name, then odsc conda publish (A) with a slug to upload to Object Storage.” B, D, and E serve other purposes—only A and C are required per OCI’s process.

Oracle Cloud Infrastructure Data Science CLI Reference, "Publishing Conda Environments".

Detailed Answer in Step-by-Step Solution:

Objective: Classify a PagerDuty notification in OCI Monitoring.

Understand OCI Monitoring: Involves events, rules, and actions for notifications.

Evaluate Options:

A: Action—executes a response (e.g., notify PagerDuty) when triggered—correct.

B: Rule—defines conditions for triggering actions—precedes the action.

C: Function—serverless code, not directly tied to notifications.

D: Event—state change triggering a rule, not the notification itself.

Reasoning: The notification is the action taken after an event/rule—fits A.

Conclusion: A is correct.

OCI documentation states: “Actions in the Monitoring service execute responses, such as sending notifications to PagerDuty, when a rule’s condition is met based on an event.” Rules (B) set conditions, Functions (C) are unrelated, and Events (D) are triggers—only Action (A) describes the notification step.

Oracle Cloud Infrastructure Monitoring Documentation, "Actions Overview".

Detailed Answer in Step-by-Step Solution:

Objective: Sequence steps for PySpark app development.

Steps:

Launch notebook: First—sets up environment.

Install PySpark conda: Second—adds Spark libraries.

Configure core-site.xml: Third—connects to data.

Develop app: Fourth—writes code.

Data Flow: Fifth—scales (optional).

Evaluate: D (1, 2, 3, 4, 5) matches this logical order.

Reasoning: Notebook first, then setup and coding.

Conclusion: D is correct.

OCI documentation states: “1) Launch a notebook session, 2) install a PySpark conda env, 3) configure core-site.xml, 4) develop your PySpark app, 5) optionally use Data Flow (D).” Other orders (A, B, C) misplace notebook launch or config—D is correct.

Oracle Cloud Infrastructure Data Science Documentation, "PySpark Development".

Detailed Answer in Step-by-Step Solution:

Objective: Enable credential-less API calls from an instance.

Understand Resource Principal: Allows instances to authenticate via IAM without user creds.

Evaluate Options:

A: Dynamic group + policy—Correct; groups instance, grants access.

B: Dynamic group only—Incomplete; needs policy.

C: User group—Irrelevant for instances.

D: No config—False; setup required.

Reasoning: A sets up resource principal fully—group and perms.

Conclusion: A is correct.

OCI documentation states: “To make API calls without credentials, create a dynamic group including the instance and add a policy (A) granting access to OCI services—enables resource principal.” B lacks policy, C is user-based, D is false—only A completes the process per OCI’s IAM setup.

Oracle Cloud Infrastructure IAM Documentation, "Resource Principal Configuration".

Detailed Answer in Step-by-Step Solution:

Objective: Get all AutoML trial results.

Understand AutoML: Trials include hyperparameter tuning outcomes.

Evaluate Options:

A: print_trials()—Displays all trial results—correct.

B: visualize_tuning_trials()—Visualizes tuning, not full list.

C: visualize_adaptive_sampling_trials()—Specific to sampling, not all trials.

D: visualize_algorithm_selection_trials()—Specific to algorithms, not all.

Reasoning: A provides comprehensive trial output.

Conclusion: A is correct.

OCI AutoML documentation states: “print_trials() outputs a table of all trials performed, including hyperparameters and scores.” Visualization methods (B, C, D) focus on specific aspects—only A gives the full list.

Oracle Cloud Infrastructure AutoML Documentation, "Trial Output Methods".

Detailed Answer in Step-by-Step Solution:

Objective: Identify encryption standard for OCI Data Science.

Understand OCI Encryption: Applies to data at rest and in transit.

Evaluate Options:

A: AES-256—Industry-standard, OCI default—correct.

B: DES—Outdated, weak—incorrect.

C: TDES—Older, less secure—incorrect.

D: Twofish—Not OCI standard—incorrect.

E: RSA—Asymmetric, not primary for data at rest—incorrect.

Reasoning: AES-256 is OCI’s go-to for Data Science resources.

Conclusion: A is correct.

OCI documentation states: “Data Science services encrypt data at rest using AES-256 (A), ensuring high security for notebooks, jobs, and models.” B, C, D, and E are either outdated or not used—only A matches OCI’s encryption policy.

Oracle Cloud Infrastructure Data Science Documentation, "Data Encryption".

Detailed Answer in Step-by-Step Solution:

Objective: Sequence steps for a PySpark app in OCI Data Science.

Evaluate Steps:

Launch notebook: First—provides the environment.

Install PySpark conda: Second—sets up Spark libraries.

Configure core-site.xml: Third—connects to data (e.g., Object Storage).

Develop app: Fourth—writes the PySpark code.

Data Flow: Fifth—optional scaling, post-development.

Check Options: D (1, 2, 3, 4, 5) matches this logical flow.

Reasoning: Notebook first, then setup, coding, and scaling.

Conclusion: D is correct.

OCI documentation recommends: “1) Launch a notebook session, 2) install a PySpark conda environment, 3) configure core-site.xml for data access, 4) develop your PySpark application, and 5) optionally use Data Flow for scale.” D follows this—others (A, B, C) misorder critical steps like launching the notebook.

Oracle Cloud Infrastructure Data Science Documentation, "PySpark in Notebooks".

Detailed Answer in Step-by-Step Solution:

Objective: Assess required skills for data scientists.

Analyze Skills:

Technical: Coding, stats, ML.

Nontechnical: Communication, business acumen.

Traits: Curiosity, problem-solving.

Reasoning: Success requires this mix—e.g., explaining models to stakeholders.

Conclusion: A (True) is correct.

OCI documentation states: “Effective data scientists combine technical skills (e.g., Python), nontechnical skills (e.g., storytelling), and traits like analytical thinking.” This holistic requirement is true (A), not false (B).

Oracle Cloud Infrastructure Data Science Documentation, "Data Scientist Skills".

Detailed Answer in Step-by-Step Solution:

Define Git: Git is a version control system—centralized vs. distributed is key.

Evaluate Options:

A: Incorrect—Git is distributed, not centralized (e.g., SVN is centralized).

B: Correct—Distributed, tracks file changes across local and remote repos.

C: Incorrect—Git allows simultaneous contributions; it manages, not prevents, merges.

D: Incorrect—Centralized is wrong, and “copious data” is vague.

Reasoning: Git’s distributed nature (each user has a full repo copy) and change-tracking are core traits.

Conclusion: B is accurate.

OCI documentation aligns with Git’s official definition: “Git is a distributed version control system that tracks changes to files, enabling collaboration and version history management.” A and D misclassify it as centralized, while C misrepresents merge handling—B captures Git’s essence as used in OCI Data Science.

Oracle Cloud Infrastructure Code Repository Documentation, "Git Overview".

Detailed Answer in Step-by-Step Solution:

Objective: Prevent deployment failure due to utilization limits.

Understand Utilization: High load requires capacity or throttling.

Evaluate Options:

A: More instances—Scales horizontally—correct.

B: Delete—Stops service, not a fix—incorrect.

C: Fewer instances—Worsens issue—incorrect.

D: Larger VM—Scales vertically—correct.

E: Reduce bandwidth—Controls load—correct.

Reasoning: A and D increase capacity, E manages demand—effective trio.

Conclusion: A, D, E are correct.

OCI documentation states: “To avoid utilization limits, increase instances (A), use a larger compute shape (D), or reduce load balancer bandwidth (E) to manage request rates.” B stops service, C reduces capacity—only A, D, E align with OCI’s deployment scaling options.

Oracle Cloud Infrastructure Data Science Documentation, "Model Deployment Scaling".

Detailed Answer in Step-by-Step Solution:

Objective: Calculate optimal bandwidth for model deployment.

Given Data:

Payload size = 1024 KB = 1024 * 8 = 8192 Kb (kilobits).

Requests/sec = 120.

Buffer = 20% = 0.2.

Calculate Base Bandwidth:

Bits/sec = Payload * Requests = 8192 Kb * 120 = 983,040 Kb/s = 983.04 Mbps.

Add Buffer:

Total = Base * (1 + Buffer) = 983.04 * 1.2 = 1179.648 Mbps.

Evaluate Options: Closest to 1179.648 Mbps is 1152 Mbps (D)—realistic rounding.

Conclusion: D is correct.

OCI documentation advises: “Calculate bandwidth as payload size (in bits) * requests/sec, then add a buffer (e.g., 20%) for peak loads.” Here, 1024 KB = 8192 Kb, * 120 = 983.04 Mbps, * 1.2 = 1179.648 Mbps. D (1152 Mbps) is the closest practical option—452 (A) and 52 (B) are too low, 7052 (C) excessive.

Oracle Cloud Infrastructure Data Science Documentation, "Model Deployment - Load Balancing".

Detailed Answer in Step-by-Step Solution:

Analyze OCI Open Data: OCI Open Data is a free service providing access to public datasets for AI/ML use cases.

Evaluate Statements:

A: True—Open Data includes text and image datasets (e.g., geospatial images).

B: False—Video and other formats may be available depending on the dataset; no strict exclusion exists.

C: False—Datasets may include metadata, but code/tooling examples aren’t guaranteed.

D: True—It’s designed for data scientists and analysts who work with datasets.

E: False—It’s not a user-contributed repository; it’s curated by Oracle.

F: False—Open Data is free and public, not subscription-based.

Select Two: A and D align with the service’s purpose and offerings.

OCI Open Data provides access to datasets like text and images (A) for AI/ML, aimed at data professionals (D). It’s a free, curated service, not user-contributed (E) or paid (F), and while it focuses on certain formats, it doesn’t explicitly exclude audio/video (B). (Reference: Oracle Cloud Infrastructure Open Data Documentation, "Overview of Open Data").

Detailed Answer in Step-by-Step Solution:

Objective: Determine where OCI budgets are set.

Understand Budgets: Track spending across OCI resources.

Evaluate Options:

A: Compartments—Scoped within tenancy, not budget root.

B: Instances—Specific resources, not budget scope.

C: Tags—Filter costs, not budget setting.

D: Tenancy—Top-level scope for budgets—correct.

Reasoning: Budgets apply at tenancy, optionally filtered (e.g., by compartment).

Conclusion: D is correct.

OCI documentation states: “Budgets are set at the tenancy level (D), with optional filters like compartments or tags to monitor spending.” A, B, and C are sub-elements—only D is the primary scope per OCI’s cost management.

Oracle Cloud Infrastructure Cost Management Documentation, "Setting Budgets".

Detailed Answer in Step-by-Step Solution:

Objective: Identify a step unique to MLOps vs. DevOps.

Compare MLOps and DevOps:

DevOps: Focuses on software deployment (CI/CD).

MLOps: Extends DevOps to ML, adding model-specific steps.

Evaluate Options:

A: Continuous deployment—Common to both (software/models).

B: Continuous integration—Common to both (code merging).

C: Continuous delivery—Common to both (releasing updates).

D: Continuous training—Unique to MLOps (retraining models with new data).

Reasoning: Only D addresses ML-specific needs (model retraining).

Conclusion: D is correct.

OCI documentation notes: “MLOps extends DevOps with continuous training, a process unique to machine learning where models are retrained with new data to maintain performance.” CI (B), CD (A), and delivery (C) are shared with DevOps—only continuous training (D) is MLOps-specific.

Oracle Cloud Infrastructure Data Science Documentation, "MLOps Concepts".

Detailed Answer in Step-by-Step Solution:

Objective: Identify ADS class for accessing datasets (e.g., scikit-learn).

Evaluate Options:

A: DatasetBrowser—Not an ADS class.

B: DatasetFactory—Loads datasets from sources like scikit-learn—correct.

C: ADSTuner—Hyperparameter tuning, not data access.

D: SecretKeeper—Manages credentials, not datasets.

Reasoning: DatasetFactory simplifies dataset loading (e.g., DatasetFactory.open()).

Conclusion: B is correct.

OCI documentation states: “DatasetFactory in ADS SDK provides methods to easily load datasets from libraries like scikit-learn or other sources (e.g., DatasetFactory.open('sklearn.datasets:load_iris')).” A isn’t real, C tunes models, and D handles secrets—only B fits.

Oracle Cloud Infrastructure ADS SDK Documentation, "DatasetFactory".

Detailed Answer in Step-by-Step Solution:

Objective: Identify the OCI service for key and secret management.

Evaluate Options:

A: Data Safe—Database security, not key management.

B: Cloud Guard—Threat detection, not keys.

C: Data Guard—DB replication, not keys.

D: Vault—Key and secret management—correct.

Reasoning: Vault is OCI’s dedicated service for crypto keys and secrets.

Conclusion: D is correct.

OCI documentation states: “OCI Vault (D) centrally manages encryption keys and secrets, securing data and resource access.” A, B, and C serve other purposes—only D matches per OCI’s securityservices.

Oracle Cloud Infrastructure Vault Documentation, "Overview".

Detailed Answer in Step-by-Step Solution:

Objective: Identify two OCI-developed, non-open-source JupyterLab extensions.

Understand Extensions: OCI enhances JupyterLab with proprietary tools.

Evaluate Options:

A: Environment Explorer—OCI-specific, non-open—correct.

B: Table of Contents—Open-source Jupyter—incorrect.

C: Command Palette—Open-source Jupyter—incorrect.

D: Notebook Examples—OCI-specific, non-open—correct.

E: Terminal—Open-source Jupyter—incorrect.

Reasoning: A and D are OCI proprietary; others are standard JupyterLab.

Conclusion: A and D are correct.

OCI documentation states: “OCI Data Science adds non-open-source extensions like Environment Explorer (A) for conda management and Notebook Examples (D) for sample code—both proprietary enhancements.” B, C, and E are open-source JupyterLab defaults—only A and D are OCI-specific per the notebook session design.

Oracle Cloud Infrastructure Data Science Documentation, "JupyterLab Extensions".

Detailed Answer in Step-by-Step Solution:

Define OCI Secrets: Secrets are sensitive data (e.g., API keys, passwords) managed securely in OCI.

Evaluate Options:

A: Object Storage is for general data, not secure secret management.

B: Vault is OCI’s service for storing and managing secrets securely.

C: Autonomous Data Warehouse is for analytics, not secret storage.

D: Oracle Databases store data, not OCI-specific secrets.

Reasoning: Vault is purpose-built for secrets with encryption and access control.

Conclusion: B is correct.

OCI Vault “provides a secure, centralized service to store and manage secrets, such as passwords and keys, with encryption at rest and fine-grained access policies.” Object Storage (A), Autonomous Data Warehouse (C), and Oracle Databases (D) serve other purposes—only Vault (B) is designed for secrets per OCI’s security architecture.

Oracle Cloud Infrastructure Vault Documentation, "Secrets Management".

Detailed Answer in Step-by-Step Solution:

Objective: Determine truth about WAF origin management.

Understand WAF: Protects apps by routing traffic via origins.

Evaluate Statements:

A: Multiple origins—True; WAF supports this.

B: Single active origin—True; only one is active per policy.

Evaluate Options:

C: B only—False; A is true.

D: Both false—Incorrect.

E: Both true—Correct per OCI WAF.

F: A only—False; B is true.

Conclusion: E is correct.

OCI documentation states: “WAF allows defining multiple origins (A), but only one origin is active per WAF policy at a time (B)—both are true (E).” C, D, and F misalign—E matches OCI’s WAF origin management.

Oracle Cloud Infrastructure WAF Documentation, "Origin Management".

Detailed Answer in Step-by-Step Solution:

Objective: Define data science’s main goal.

Evaluate Options:

A: Archiving—Not the focus; too narrow.

B: Analyze for insights/business value—Core purpose—correct.

C: Prep for analytics—Means, not the end goal.

D: Output-focused—Vague, incomplete.

Reasoning: B captures the actionable insight generation central to data science.

Conclusion: B is correct.

OCI documentation defines data science as “mining and analyzing large datasets to uncoveractionable insights for operational improvements and business value.” A is storage-focused, C is preparatory, and D is unclear—only B reflects the principal goal per OCI’s mission.

Oracle Cloud Infrastructure Data Science Documentation, "What is Data Science?".

Detailed Answer in Step-by-Step Solution:

Objective: Scale up notebook without losing work.

Evaluate Options:

A: Object Storage—Extra steps, inefficient.

B: Block volume—Persists data, seamless scale—correct.

C: Local machine—Risky, cumbersome.

D: Recreate—Loses work, impractical.

Reasoning: B uses OCI’s built-in persistence.

Conclusion: B is correct.

OCI documentation states: “Files in /home/datascience (B) persist on block volume; deactivate, then reactivate with a larger shape to scale up without data loss.” A, C, and D add complexity or risk—only B is optimal per OCI’s design.

Oracle Cloud Infrastructure Data Science Documentation, "Scaling Notebook Sessions".

Detailed Answer in Step-by-Step Solution:

Objective: Determine which validator handler runs for is_tech_job(job_family='Engineering').

Understand Validator System: Likely ADS SDK—executes handlers based on conditions.

Analyze Validators:

Default: is_tech_job_default_handler—No condition, fallback.

Open: is_tech_job_open_handler—Condition ('job_family',)—requires job_family arg.

Closed: is_tech_job_closed_handler—Condition ('job_family': 'IT')—requires job_family='IT'.

Evaluate Call: job_family='Engineering'—Matches job_family presence, not IT.

Reasoning:

Open handler applies (tuple condition means arg exists).

Closed fails (Engineering ≠ IT).

Default is overridden by specific matches.

Conclusion: D is correct.

OCI ADS documentation states: “Validators execute the most specific handler matching the condition; for is_tech_job(job_family='Engineering'), is_tech_job_open_handler (D) runs as it matches job_family presence, while is_tech_job_closed_handler (C) requires IT—default (A) is bypassed, no error (B).” Only D fits per ADS validator logic.

Oracle Cloud Infrastructure ADS SDK Documentation, "Feature Type Validators".

Detailed Answer in Step-by-Step Solution:

Objective: Identify three data formats supported by ADS SDK for loading data.

Understand ADS SDK: Facilitates data loading into notebook sessions via DatasetFactory.

Evaluate Options:

A. DOCX: Not natively supported—requires conversion (e.g., to text).

B. Pandas DataFrame: Supported—core format for data manipulation in ADS.

C. JSON: Supported—common structured data format.

D. Raw Images: Not directly supported—image data needs preprocessing (e.g., via Vision).

E. XML: Supported—parseable structured format.

Reasoning: ADS focuses on tabular/structured data—B, C, E align; A and D require external handling.

Conclusion: B, C, E are correct.

OCI documentation states: “ADS SDK’s DatasetFactory supports loading data from formats like Pandas DataFrames (B), JSON (C), and XML (E), enabling easy integration into notebook sessions.” DOCX (A) isn’t natively handled, and raw images (D) require preprocessing outside ADS—B, C, E match the supported list.

Oracle Cloud Infrastructure ADS SDK Documentation, "Supported Data Formats".

Detailed Answer in Step-by-Step Solution:

Objective: Define the role of score.py in OCI model artifacts.

Understand Artifacts: score.py is key for deployment runtime.

Evaluate Options:

A: Infra config—Handled by OCI settings, not score.py.

B: Inference logic—Correct; runs load_model(), predict().

C: Scaling—Set in deployment, not score.py.

D: Dependencies—In runtime.yaml, not score.py.

Reasoning: B aligns with score.py’s execution role.

Conclusion: B is correct.

OCI documentation states: “score.py (B) contains the inference logic, including functions to load the model and predict outputs, executed by the deployment endpoint.” A, C, and D are managed elsewhere—only B matches OCI’s design.

Oracle Cloud Infrastructure Data Science Documentation, "Model Artifact - score.py".

Detailed Answer in Step-by-Step Solution:

Understand OCI Data Science Pipelines: Pipelines automate ML workflows with configurable steps.

Check Override Mechanism: Defaults (e.g., compute shape, storage) can be modified before execution via the OCI Console, SDK, or CLI.

Evaluate Options:

A: False—Overrides can occur post-creation, before running.

B: False—Any authorized user, not just admins, can override defaults.

C: True—Settings can be adjusted before execution starts.

D: False—Defaults can be changed post-creation, pre-execution.

Conclusion: C is correct as it reflects the flexibility of pipeline configuration.

OCI Data Science Pipelines allow users to override default settings (e.g., compute resources, environment variables) before execution, as noted in the official documentation. This can be done via the UI or programmatically, offering flexibility beyond creation time (A) and without admin-only restrictions (B). (Reference: Oracle Cloud Infrastructure Data Science Pipelines Documentation, "Configuring Pipelines").

Detailed Answer in Step-by-Step Solution:

Objective: Identify three data types for OCI Data Labeling (question likely incomplete—assuming B, C, D, E options).

Understand Data Labeling: Annotates data for ML—focuses on specific types.

Evaluate Options (Assuming Typical Set):

A: Audio—Not supported—incorrect.

B: Text Document—Supported (e.g., NER)—correct.

C: Images—Supported (e.g., object detection)—correct.

D: Graphs—Not a standard type—incorrect.

Assumed E: Videos—Supported but missing—adjust to fit.

Reasoning: OCI supports text, images, and videos—question lists only four, so B and C are definite.

Conclusion: B, C (third likely video, missing).

OCI documentation states: “Data Labeling supports text documents (B), images (C), and videos for annotation—audio (A) and graphs (D) are not included.” Question likely meant three from a larger set; B and C are confirmed per OCI’s Data Labeling capabilities.

Oracle Cloud Infrastructure Data Labeling Documentation, "Supported Data Types".

Detailed Answer in Step-by-Step Solution:

Objective: Define LIME’s core concept.

Understand LIME: Explains individual predictions with local surrogate models.

Evaluate Options:

A: Complex global, simple local—Correct LIME principle.

B: Agnosticism—True but not the key idea.

C: Global/local similarity—False.

D: Local vs. global agnosticism—Incorrect distinction.

Reasoning: A captures LIME’s local approximation focus.

Conclusion: A is correct.

OCI documentation notes: “LIME (A) explains predictions by approximating complex global models with simpler local surrogate models around specific instances.” B, C, and D misalign—only A reflects LIME’s foundational idea per OCI’s interpretability tools.

Oracle Cloud Infrastructure Data Science Documentation, "Model Interpretability - LIME".

Detailed Answer in Step-by-Step Solution:

Objective: Identify three stages in OCI AutoML Pipeline.

Understand Pipeline: Automates ML steps from data to model training.

Evaluate Options:

A: Feature Selection—Selects relevant features—correct.

B: Adaptive Sampling—Reduces data size—correct.

C: Model Deployment—Post-pipeline step—incorrect.

D: Feature Extraction—Not explicit in OCI AutoML—incorrect.

E: Algorithm Selection—Chooses best model—correct.

Reasoning: A, B, E are core automated stages; C and D are separate.

Conclusion: A, B, E are correct.

OCI documentation lists “AutoML Pipeline stages as adaptive sampling (B), feature selection (A), algorithm selection (E), and hyperparameter tuning.” Deployment (C) is post-pipeline, and extraction (D) isn’t highlighted—only A, B, E are included per OCI’s design.

Oracle Cloud Infrastructure AutoML Documentation, "Pipeline Components".

Detailed Answer in Step-by-Step Solution:

Understand OCI Data Labeling Capabilities: OCI Data Labeling is designed to annotate data for machine learning, supporting specific file types like images, text documents, and videos.

Evaluate Options:

A. Video footage: Supported for tasks like object detection or action recognition.

B. Images: Supported for image classification, object detection, etc.

C. Typewritten document: Supported as text data for tasks like entity extraction or classification.

D. Purchase orders: While potentially text-based, this is ambiguous without format clarification (e.g., PDF, image). OCI supports text annotation, but “purchase orders” isn’t a specific file type—it’s assumed as text here.

E. Audio recording: Not supported, as OCI Data Labeling focuses on visual and textual data, not audio.

Select Three: A (video), B (images), and C (text documents) are explicitly supported file types.

OCI Data Labeling supports annotating datasets of images, text, and videos, as per the official documentation. Video footage (A) can be annotated for tasks like object tracking, images (B) for classification or detection, and typewritten documents (C) for text-based annotations (e.g., named entity recognition). Audio files (E) are not supported, and while purchase orders (D) could be text, the question specifies “typewritten document” as a clearer match. (Reference: Oracle Cloud Infrastructure Data Labeling Service Documentation, "Supported Data Types").

Detailed Answer in Step-by-Step Solution:

Objective: Identify interfaces to save a 2 GB model to the Model Catalog.

Evaluate Options:

A: OCI CLI—Supports Data Science tasks—possible but not primary.

B: ADS SDK—Designed for model catalog ops—correct.

C: ODSC CLI—Not standard; likely typo for OCI CLI.

D: Console—GUI for catalog uploads—correct.

E: OCI Python SDK—Programmatic catalog access—correct.

F: Git CLI—Version control, not catalog-related.

Reasoning: B, D, E are OCI’s primary interfaces; A is valid but less emphasized.

Conclusion: B, D, E are correct (A plausible but not top-tier).

OCI documentation lists “ADS SDK (B), OCI Console (D), and OCI Python SDK (E) as primary methods to save models to the Model Catalog.” OCI CLI (A) works but isn’t highlighted, C isn’t real, and F is unrelated—B, D, E are the standard trio.

Oracle Cloud Infrastructure Data Science Documentation, "Model Catalog Interfaces".

Detailed Answer in Step-by-Step Solution:

Objective: Optimize an embarrassingly parallel job in OCI Data Science Jobs.

Define Embarrassingly Parallel: Tasks are independent, ideal for simultaneous runs.

Evaluate Options:

A: Multiple simultaneous runs—Leverages parallelism—correct.

B: One job per run—Misstates capability; unnecessary complexity.

C: Sequential runs—Inefficient, ignores parallelism.

D: False—Jobs support parallelism.

Reasoning: A maximizes efficiency for parallel tasks.

Conclusion: A is correct.

OCI documentation states: “For embarrassingly parallel workloads, create a single Job and launch multiple simultaneous Job Runs to process data in parallel.” B misinterprets limits, C wastes time, and D denies capability—only A fits OCI’s design.

Oracle Cloud Infrastructure Data Science Documentation, "Parallel Job Runs".

Detailed Answer in Step-by-Step Solution:

Understand OCI Anomaly Detection: This service trains models to detect anomalies in multivariate data (e.g., sensor readings), requiring data to be accessible within OCI’s ecosystem.

Assess Storage Requirements: The training data must be in a scalable, OCI-compatible location that the Anomaly Detection service can access programmatically.

Evaluate Options:

A. Your local machine: Data on a local machine isn’t directly accessible to OCI services without upload, making it impractical for cloud-based training.

B. MySQL database: While OCI supports MySQL, Anomaly Detection doesn’t natively integrate with it for training data; it prefers file-based input.

C. Autonomous Data Warehouse: This is a database for analytics, not the default storage for Anomaly Detection training data, which expects CSV/JSON files.

D. Object Storage Bucket: OCI Object Storage is a scalable, durable storage service that Anomaly Detection uses to ingest training data (e.g., CSV files).

Reasoning: Object Storage is the standard for large-scale data in OCI services, offering seamless integration with Anomaly Detection via APIs or SDKs.

Conclusion: D is the correct choice as it aligns with the service’s architecture.

The OCI Anomaly Detection service requires training data to be uploaded to an Object Storage Bucket in formats like CSV or JSON. This is explicitly outlined in the official documentation, which states that users must “upload the training dataset to an OCI Object Storage bucket” before creating a data asset for model training. Local storage (A) isn’t viable for cloud processing, and databases like MySQL (B) or Autonomous Data Warehouse (C) aren’t supported as primary inputs. Object Storage (D) provides the scalability and accessibility needed for multivariate anomaly detection workflows.

Oracle Cloud Infrastructure Anomaly Detection Documentation, "Preparing Training Data" section.

Detailed Answer in Step-by-Step Solution:

Objective: Identify IAM components in OCI.

Understand IAM: Manages access via policies, groups, etc.

Evaluate Options:

A: Policies—Core IAM component—correct.

B: Subnets—Networking, not IAM.

C: Instances—Compute, not IAM.

D: VCN—Networking, not IAM.

Reasoning: Only A is an IAM element—others are resources.

Conclusion: A is correct.

OCI documentation states: “IAM includes components like policies (A), groups, and compartments to control resource access.” B, C, and D are infrastructure, not IAM—only A fits per OCI’s IAM framework.

Oracle Cloud Infrastructure IAM Documentation, "IAM Components".

Detailed Answer in Step-by-Step Solution:

Objective: Configure conda env for a pipeline step.

Evaluate Options:

A: Shape—Infra, not env config.

B: Volume—Storage, not env.

C: Command-line—Step args, not env.

D: Env variables—Sets conda path—correct.

Reasoning: D specifies runtime env (e.g., CONDA_ENV_SLUG).

Conclusion: D is correct.

OCI documentation states: “Configure a pipeline step’s conda environment using environment variables (D), such as CONDA_ENV_SLUG, in the step definition.” A, B, and C address other aspects—only D fits env config.

Oracle Cloud Infrastructure Data Science Documentation, "Pipeline Step Configuration".

Detailed Answer in Step-by-Step Solution:

Objective: Locate sources for ADS SDK in OCI.

Understand ADS SDK: A Python library for Data Science tasks (e.g., EDA).

Evaluate Options:

A: Big Data Service—Spark-focused, not ADS source.

B: Machine Learning—Separate service, not ADS-related.

C: Conda in OCI Data Science—Preinstalled ADS in notebook sessions.

D: PyPI—Public source to install ADS (pip install oracle-ads).

E: ADW—Database, not an SDK source.

Reasoning: C (preinstalled) and D (installable) are practical access points.

Conclusion: C and D are correct.

OCI documentation states: “The ADS SDK is available in OCI Data Science notebook sessions via preinstalled conda environments (C) and can be installed from PyPI (D) using pip install oracle-ads.” Big Data (A), Machine Learning (B), and ADW (E) don’t host ADS—only C and D apply.

Oracle Cloud Infrastructure Data Science Documentation, "ADS SDK Installation".

Detailed Answer in Step-by-Step Solution:

Understand Model Catalog: The OCI Model Catalog stores trained models and associated details.

Identify Shareable Artifacts: When sharing, all components—model file, metadata (e.g., name, description), hyperparameters (e.g., learning rate), and metrics (e.g., accuracy)—are accessible.

Evaluate Options:

A: Excludes the model itself—incorrect.

B: Excludes metrics—incorrect.

C: Excludes metadata and hyperparameters—incorrect.

D: Includes all components—correct.

Conclusion: D is comprehensive and accurate.

The OCI Model Catalog allows sharing of the model artifact (the trained model), metadata, hyperparameters, and performance metrics, enabling full examination by colleagues. This isdetailed in the official documentation. (Reference: Oracle Cloud Infrastructure Data Science Documentation, "Model Catalog Overview").

Detailed Answer in Step-by-Step Solution:

Context Analysis: You need to investigate an existing deep neural network model in the OCI Model Catalog with no prior information.

Understand Model Catalog: The Model Catalog stores trained models along with metadata, hyperparameters, and provenance (origin and history) details.

Evaluate Options:

A. Refer to the code inside the model: The model artifact (e.g., a serialized file like .pkl) doesn’t typically include readable source code; it’s a trained object, not the training script.

B. Check for model taxonomy details: Taxonomy (e.g., classification vs. regression) provides high-level categorization but lacks specifics like framework or architecture.

C. Check for metadata tags: Metadata includes name, description, and tags, offering some context but not detailed framework info (e.g., TensorFlow vs. PyTorch).

D. Check for provenance details: Provenance tracks the model’s creation process, including the framework, training environment, and data sources, providing the most comprehensive insight.

Reasoning: Provenance details are designed to document the “how” and “what” of model creation, making them ideal for uncovering the framework (e.g., Keras, PyTorch) and other specifics absent from initial handover.

Conclusion: D is the best approach for detailed investigation.

In OCI Data Science, the Model Catalog stores provenance information, which includes “details about the model’s origin, such as the framework used (e.g., TensorFlow, PyTorch), the training environment, and dataset references.” This is more informative than metadata tags (C), which are user-defined and less structured, or taxonomy (B), which is broad. The model artifact (A) is a binary file (e.g., pickle), not a readable codebase. Provenance (D) offers a detailed audit trail, critical for analyzing an undocumented deep neural network model like this one.

Oracle Cloud Infrastructure Data Science Documentation, "Model Catalog - Provenance Details" section.

Detailed Answer in Step-by-Step Solution:

Understand Notebook Sessions: These are OCI compute instances running JupyterLab.

Deactivation Impact: Deactivating stops the session to save costs.

Evaluate Options:

A: False—Costs decrease as compute stops.

B: False—Boot volume data isn’t preserved; block volume data is.

C: True—The compute instance shuts down, halting billing.

D: False—Block volume persists unless explicitly deleted.

Reasoning: Deactivation stops the instance (C), preserving block volume data separately.

Conclusion: C is correct.

The OCI documentation states: “When a notebook session is deactivated, the underlying compute instance stops, and billing for compute resources ceases. Data on the attached block volume is preserved, but the boot volume is not.” A is backwards, B misattributes preservation, and D overstates deletion—only C aligns with the process.

Oracle Cloud Infrastructure Data Science Documentation, "Notebook Session Lifecycle".

Detailed Answer in Step-by-Step Solution:

Objective: Find optimal compute shape balancing cost and time.

Approach: Iterative testing with metrics (e.g., CPU/memory usage, runtime).

Evaluate Options:

A: Tuning parameters when underutilized—focuses on model, not shape optimization.

B: Strongest shape—Costly, ignores balance; overkill likely.

C: Scale up from small shape when fully utilized—Balances cost/time effectively.

D: Random start with pre-tests—Unsystematic and inefficient.

Reasoning: C incrementally increases resources based on utilization, optimizing both factors.

Conclusion: C is correct.

OCI documentation advises: “To optimize compute shape for Jobs, start with a small shape, monitor utilization (e.g., CPU, memory) and runtime via OCI Monitoring. If fully utilized, scale up until performance plateaus—balancing cost and speed.” A misfocuses on model tuning, B wastes cost, and D lacks structure—only C aligns with this method.

Oracle Cloud Infrastructure Data Science Documentation, "Optimizing ComputeShapes for Jobs".