Machine Learning. Test

Is the following statement True or False?

The supervised, unsupervised, and reinforcement learning paradigms within machine learning do not apply to deep learning.

Is the following statement True or False?

An AI can only be trained under one of these three paradigms (supervised, unsupervised, and reinforcement learning), so they cannot overlap.

Is the following statement True or False?

At its core, machine learning is really about using math, statistics, and algorithms to find patterns and make predictions — whether it’s recognising music styles, grouping similar songs, or predicting the next data point.

Match each machine learning paradigm with its correct description.

Read the case and select the type of learning corresponding to the statement.

A music recommendation system is trained using a dataset where human experts manually label each song with its correct genre (e.g., rock, jazz, electronic). The AI learns to classify new songs based on these labels.

Read the case and select the type of learning corresponding to the statement.

A music streaming platform groups songs with similar sonic characteristics (e.g., tempo, instrumentation, harmony) together without human-labeled data. The AI naturally clusters related tracks, helping users discover similar music.

Select the type of learning corresponding to the statement.

A DJ software observes real-time transitions made by professional DJs. It receives feedback from the DJs (e.g., a rating of how smooth the transition was) and adjusts its future recommendations accordingly.

Select the type of learning corresponding to the statement:

A music production AI is trained using thousands of drum patterns where human producers have explicitly labeled each as “jazz,” “hip-hop,” “rock,” etc. The AI learns to generate beats that match these styles.

Select the type of learning corresponding to the statement:

A record label uses AI to analyze thousands of new artists’ songs, clustering them based on vocal timbre, instrumentation, and lyrical themes—without predefined categories or human input.

Select the type of learning corresponding to the statement:

An AI-powered mastering tool is trained on a dataset where sound engineers have labeled tracks as “well-mastered” or “poorly-mastered.” The AI learns to enhance audio quality based on these human-labeled examples.

Select the type of learning corresponding to the statement:

A virtual drummer generates drum fills for a song and gets rated by human producers. Over time, it adjusts its style to get higher ratings and avoid mistakes, improving its performance dynamically.

Select the type of learning corresponding to the statement:

A playlist generator groups songs into mood-based categories (e.g., “chill,” “energetic”) based solely on the song’s characteristics — without using human-labeled mood tags.

Select the type of learning corresponding to the statement:

An AI composer generates melodies and gets real-time feedback on whether they match a desired emotion (e.g., happy, sad). It modifies its compositions based on this feedback, improving over time.

Select the type of learning corresponding to the statement:

A copyright detection system is trained on a dataset where human experts label whether a song contains unauthorized samples or not. The AI learns to detect potential copyright infringements.