/en-tutor-prep-s5-ict-artificial-intelligence-machine-learning-practice-1/
/en-tutor-prep-s5-ict-artificial-intelligence-machine-learning-practice-1/
General Inquiry: info@TutorZone.com.hk
Office Hours: Monday to Saturday - 9:00 AM to 6:00 PM
WhatsApp Monday to Sunday - 24 hours
Hotline: +852 6828 1809
Instant Support: +852 9061 3106

S5 ICT: AI and machine learning basics, ten items

S5 ICT: AI and machine learning basics, ten items

  • 2026-11-20

S5 ICT: AI and machine learning basics, ten items

Paper one builds the concepts: how AI, machine learning and deep learning nest, supervised against unsupervised learning, splitting data into training and test sets, classification against regression, accuracy and the confusion matrix, overfitting and the weight count of a network.

Foundation: concepts, data splits and three measures This paper starts from definitions: the scope narrows from AI to machine learning to deep learning. Supervised learning relies on labelled data while unsupervised learning has no labels. Of 500 photos, 80 percent or 400 go to training and 100 to testing; of 1 200 samples, 70 percent or 840 train and 360 test. Answering 45 of 50 cases correctly is 90 percent accuracy. The confusion matrix totals 30 and both recall and accuracy come out at 80 percent. A categorical output means classification and a continuous one means regression. Training at 99 against testing at 62 is a 37 point gap, which is overfitting. Three by four gives twelve weight connections. One thousand 64 by 64 colour images take 12 288 000 bytes, about 11.72 MB.

01 Paper details (check answers afterwards)

Item Detail
Questions 10
Suggested time 25
Difficulty mix Foundation 3 · Intermediate 3 · Challenge 3 · Error redo 1
Chapter Chapter: S5 ICT: Artificial Intelligence and Machine Learning
Level Foundation (10 items)

02 Questions (10 items)

  1. FoundationHow are artificial intelligence, machine learning and deep learning related?
  2. FoundationWhat is the key difference between supervised and unsupervised learning?
  3. FoundationOf 500 photos, 80 percent go to the training set and the rest to the test set. How many in each?
  4. IntermediateWith 1 200 labelled samples and 70 percent used for training, how large is the test set?
  5. IntermediateA model answers 45 of 50 test cases correctly. What is its accuracy?
  6. IntermediateA confusion matrix shows 12 hits, 3 misses, 3 false alarms and 12 correct rejections. What are recall and accuracy?
  7. ChallengePredicting pass or fail, and predicting a score from 0 to 100, belong to which task types?
  8. ChallengeTraining accuracy is 99 percent but test accuracy is only 62 percent. What is wrong and how can it be improved?
  9. ChallengeAn input layer has 3 neurons and a hidden layer has 4. How many weight connections join the two layers?
  10. Error redoA training set holds 1 000 colour images of 64 by 64 pixels at 3 bytes per pixel. What is the size in bytes and in MB?

03 Answers and one-line notes

# Answer One-line note
1 Machine learning is part of AI, and deep learning is a machine learning method that uses multi-layer neural networks. The scope narrows from AI to machine learning to deep learning, so the three are not parallel terms.
2 Supervised learning trains on labelled data that carries the correct answer, while unsupervised learning uses unlabelled data. Labels allow an error to be measured and the model adjusted; without labels the model looks for clusters or patterns instead.
3 400 in the training set and 100 in the test set. 500 times 0.8 gives 400, and the remaining 500 minus 400 leaves 100.
4 840 for training and 360 for testing. 1 200 times 7 divided by 10 is 840, and 1 200 minus 840 leaves 360.
5 90 percent. 45 divided by 50 is 0.9, which is 90 percent.
6 Both are 80 percent. There are 30 cases; recall is 12 over 15 and accuracy is 24 over 30, so both give 80 percent here.
7 The first is classification and the second is regression. A categorical label as output means classification; a continuous number as output means regression.
8 This is overfitting; add more training data, simplify the model or strengthen regularisation. The 37 point gap shows the model memorised the training samples instead of learning the pattern.
9 12 connections. In a fully connected layer every input neuron reaches every hidden neuron, so 3 times 4 gives 12.
10 12 288 000 bytes, about 11.72 MB. 1 000 times 64 times 64 times 3 gives 12 288 000, and dividing by 1 048 576 gives 11.72 MB.

04 What to do after marking

  • After marking, redo only the wrong items and write down which step went wrong (formula, unit, or misread number).
  • Repeat the same item two days later; move to the next paper only when both attempts are correct.
  • Parents only need the answer and note columns; there is no need to explain every item.

Want to try a train and test split yourself? Book a TutorZone tutor and check all ten items in 25 minutes.Tutor Registration →

Questions and answers are written originally by the TutorZone team and are not taken from any school, publisher or examination authority paper. Every value is checked by computer.

Join as a tutor or coach — academic, music, sports, arts, STEM. Free to register. See how it works →