Machine learning vs rule-based systems in AI for Everyone - Visual Side-by-Side Comparison

Concept Flow - Machine learning vs rule-based systems

Start: Input Data

↓

Apply Rules

↓

Output Result

↓

End: Decision Made

The system receives input data, then either applies fixed rules or uses a trained model to produce an output.

Execution Sample

AI for Everyone

Input: data
If rule-based:
  Apply fixed rules
Else:
  Use learned model
Output result

This pseudocode shows how input data is processed differently by rule-based and machine learning systems.

Analysis Table

Step	Input	System Type	Action	Output
1	Temperature=30	Rule-based	Check if temp > 25	Output: 'Hot'
2	Temperature=20	Rule-based	Check if temp > 25	Output: 'Not Hot'
3	Temperature=30	Machine learning	Model predicts based on data	Output: 'Likely Hot'
4	Temperature=20	Machine learning	Model predicts based on data	Output: 'Likely Not Hot'
5	Temperature=15	Rule-based	Check if temp > 25	Output: 'Not Hot'
6	Temperature=15	Machine learning	Model predicts based on data	Output: 'Likely Not Hot'
7	-	-	-	End of processing

💡 Processing ends after all inputs are evaluated by both systems.

State Tracker

Variable	Start	After 1	After 2	After 3	After 4	After 5	After 6	Final
Input Temperature	N/A	30	20	30	20	15	15	N/A
Rule-based Output	N/A	'Hot'	'Not Hot'	N/A	N/A	'Not Hot'	N/A	N/A
ML Output	N/A	N/A	N/A	'Likely Hot'	'Likely Not Hot'	N/A	'Likely Not Hot'	N/A

Key Insights - 3 Insights

Why does the rule-based system give exact outputs while machine learning gives 'likely' outputs?

Rule-based systems follow fixed rules (see steps 1,2,5 in execution_table) so outputs are certain. Machine learning predicts based on patterns and probabilities (steps 3,4,6), so outputs are probabilistic.

Can the rule-based system learn from new data during execution?

No, rule-based systems do not learn during execution; they only apply predefined rules as shown in the execution_table steps 1,2,5.

Why do both systems process the same input differently?

Because rule-based systems use fixed conditions, while machine learning models use patterns learned from data, leading to different outputs for the same input (compare steps 1 and 3).

Visual Quiz - 3 Questions

Look at the execution_table at step 2. What output does the rule-based system produce for temperature 20?

A 'Hot'

B 'Not Hot'

C 'Likely Hot'

D 'Likely Not Hot'

💡 Hint

Check the 'Rule-based Output' column in variable_tracker after step 2.

At which step does the machine learning system predict 'Likely Hot'?

A Step 1

B Step 3

C Step 5

D Step 6

💡 Hint

Look at the 'ML Output' column in variable_tracker and match with execution_table steps.

If the rule in the rule-based system changed to temp > 20, what would be the output at step 2?

A 'Hot'

B 'Not Hot'

C 'Likely Hot'

D 'Likely Not Hot'

💡 Hint

Refer to step 2 in execution_table and consider the condition change.

Concept Snapshot

Machine learning systems learn patterns from data to predict outputs.
Rule-based systems use fixed rules to decide outputs.
Rule-based outputs are exact; ML outputs are probabilistic.
ML adapts with data; rule-based does not change unless rules are updated.
Both process inputs but differ in flexibility and learning ability.

Full Transcript

This visual execution compares machine learning and rule-based systems. Input data is processed either by applying fixed rules or by using a trained model. The execution table shows step-by-step how each system handles temperature inputs, producing outputs like 'Hot' or 'Likely Hot'. Variables track input values and outputs over steps. Key moments clarify why rule-based outputs are exact and machine learning outputs are probabilistic. The quiz tests understanding of outputs at specific steps and effects of changing rules. The snapshot summarizes the main differences: rule-based systems use fixed rules, machine learning systems learn from data and predict probabilistically.

Practice

1. Which of the following best describes a machine learning system compared to a rule-based system?

easy

A. It only works with simple, clear instructions.

B. It follows fixed rules without change.

C. It learns from data and adapts over time.

D. It cannot improve after deployment.

Solution

1. Step 1: Understand machine learning characteristics

   Machine learning systems learn from examples and improve with more data.

2. Step 2: Compare with rule-based systems

   Rule-based systems follow fixed instructions and do not adapt.

3. Final Answer:

   It learns from data and adapts over time. -> Option C

4. Quick Check:

   Machine learning = adapts [OK]

Hint: Machine learning adapts; rule-based does not [OK]

Common Mistakes:

• Confusing fixed rules with learning
• Thinking rule-based systems adapt
• Assuming machine learning cannot improve

2. Which syntax correctly describes a rule-based system?

easy

A. train_model(data) to predict temperature

B. if temperature > 30 then turn_on_fan() else turn_off_fan()

C. learn_from_data(data) to adjust fan speed

D. update_rules_based_on_feedback()

Solution

1. Step 1: Identify rule-based syntax

   Rule-based systems use fixed if-then rules like 'if temperature > 30 then turn_on_fan()'.

2. Step 2: Check other options

   Options A, C, and D describe learning or updating, which are machine learning concepts.

3. Final Answer:

   if temperature > 30 then turn_on_fan() else turn_off_fan() -> Option B

4. Quick Check:

   Rule-based = fixed if-then rules [OK]

Hint: Rule-based uses fixed if-then rules [OK]

Common Mistakes:

• Confusing learning functions with rules
• Choosing options that imply adaptation
• Ignoring fixed condition-action format

3. Consider this simple system:
rules = {'hot': 'turn_on_ac', 'cold': 'turn_on_heater'}
def apply_rule(temp):
    if temp > 25:
        return rules['hot']
    else:
        return rules['cold']
print(apply_rule(30))
What will this print?

medium

A. Error

B. turn_on_heater

C. null

D. turn_on_ac

Solution

1. Step 1: Analyze the input and condition

   Input temperature is 30, which is greater than 25, so the 'hot' rule applies.

2. Step 2: Determine the returned action

   The function returns rules['hot'], which is 'turn_on_ac'.

3. Final Answer:

   turn_on_ac -> Option D

4. Quick Check:

   Temp 30 > 25 -> 'turn_on_ac' [OK]

Hint: Check condition then pick matching rule [OK]

Common Mistakes:

• Choosing 'turn_on_heater' ignoring condition
• Assuming function returns null
• Thinking code causes error

4. This code tries to use a rule-based system but has a bug:
rules = {'hot': 'turn_on_ac', 'cold': 'turn_on_heater'}
def apply_rule(temp):
    if temp > 25:
        return rules['hot']
    elif temp <= 25:
        return rules['cold']
print(apply_rule(25))
What is the bug and how to fix it?

medium

A. Bug: 'elif' should be 'else'; fix by replacing 'elif' with 'else'.

B. Bug: Missing rule for temp=25; fix by adding 'temp == 25' rule.

C. Bug: KeyError on 'cold'; fix by adding 'cold' key to rules.

D. Bug: Function does not return anything; fix by adding return statement.

Solution

1. Step 1: Identify condition overlap

   The code uses 'elif temp <= 25', but temp=25 matches this condition. However, since 'if temp > 25' fails implies 'temp <= 25', the elif is redundant.

2. Step 2: Check if 'elif' is necessary

   Since the first condition is 'temp > 25', the else branch can cover all other cases, so 'else' is simpler and clearer.

3. Final Answer:

   Bug: 'elif' should be 'else'; fix by replacing 'elif' with 'else'. -> Option A

4. Quick Check:

   Use else for remaining cases [OK]

Hint: Use else for all other cases, not elif [OK]

Common Mistakes:

• Thinking temp=25 is missing
• Assuming KeyError occurs
• Believing function lacks return

5. You want to build a system that detects spam emails. The rules for spam change often and new patterns appear regularly. Which approach is best and why?

hard

A. Use machine learning because it can learn new spam patterns from data.

B. Use a rule-based system because rules are easy to write and fixed.

C. Use a rule-based system because it never makes mistakes.

D. Use machine learning because it requires no data to work.

Solution

1. Step 1: Understand problem requirements

   Spam patterns change often, so fixed rules will become outdated quickly.

2. Step 2: Choose approach based on adaptability

   Machine learning can learn from new data and adapt to new spam patterns automatically.

3. Final Answer:

   Use machine learning because it can learn new spam patterns from data. -> Option A

4. Quick Check:

   Changing patterns = machine learning [OK]

Hint: Changing rules? Choose machine learning [OK]

Common Mistakes:

• Choosing rule-based for changing patterns
• Thinking machine learning needs no data
• Assuming rule-based systems never err