Understanding Fuzzy Logic: Making Computers Grasp the Nuances

Brian Chan
4 min readSep 27, 2023

--

Have you ever described something as “kind of hot,” “quite hot,” or “very hot”? It’s something we do all the time as humans. However, when it comes to computers, they struggle to grasp these fuzzy descriptors. They need clear, quantifiable information to work with. This is where fuzzy logic steps in to bridge the gap.

In standard logic, computers operate in a binary world of 1s and 0s, true or false. For instance, you might define anything above 40°C as “hot.” But what about something at 39°C? Is it not hot at all? Fuzzy logic comes to the rescue.

Fuzzy logic introduces the concept of degree of truth, replacing the strict 1s and 0s with a more nuanced representation. For example, imagine a cup of water at 40°C being “0.9 hot,” while anything at 43°C or higher is “1 hot,” and anything at 35°C or lower is “0 hot.”

In fuzzy logic, “hot” isn’t just a single value; it’s a range of values between 0 and 1. It’s essential to note that the degree of truth isn’t the same as probability. It measures the uncertainty about a statement, not the likelihood.

With fuzzy logic, computers can now understand the richness of human language and the nuances it carries.

Membership Functions

So, how does fuzzy logic work its magic? It uses something called membership functions. These functions map input values to degrees of truth. In the earlier example, there were two membership functions. The one on the right says that at 40, the degree of truth is 0.9, while at 30, it’s 0.

There are various common membership functions, such as triangular and trapezoidal, each serving different purposes.

Rules

In fuzzy logic, we rely on IF-THEN rules to create conditions. These rules trigger different outputs based on whether the input values satisfy these conditions. The more complex the system, the more rules you might need.

Here’s an example from Wikipedia:

IF Temperature IS Very Cold THEN Stop the Fan

IF Temperature IS Cold THEN Slow Down the Fan

IF Temperature IS Normal THEN Maintain the Current Level

IF Temperature IS Hot THEN Speed Up the Fan

In this example, “Very Cold,” “Cold,” “Normal,” and “Hot” each have their corresponding Membership Functions. If an input temperature maps to a degree of truth other than 0 in any of these functions, the associated rule is triggered. For instance, if the temperature is 26°C, it’s 0.8 “Normal” and 0.3 “Hot,” so both the second and third rules are triggered simultaneously.

Defuzzification

In fuzzy logic, we use membership functions and IF/THEN rules to determine the degree of truth for different conditions. However, to make a practical decision or control an action, we need to convert these fuzzy inputs into clear, actionable outputs. This process is known as defuzzification.

In our example, imagine we have measured the temperature, and it falls into two categories: “Normal” and “Hot.” The degrees of truth for these categories are 0.8 and 0.3, respectively.

To decide how to control the fan, we need to find a single, crisp value that represents the appropriate fan speed. There are various methods for defuzzification, but one common approach is the centroid method.

The centroid method calculates the center of mass of the areas under the membership functions that were activated. In our case, the areas under “Normal” and “Hot” membership functions are activated with degrees of truth 0.8 and 0.3, respectively.

Here’s how the calculation works:

  1. Calculate the weighted average of the activated areas. This involves multiplying each degree of truth by the position of its centroid and then dividing by the sum of the degrees of truth.
(0.8 * x1 + 0.3 * x2) / (0.8 + 0.3)

Where:

  • x1 is the centroid of the “Normal” membership function.
  • x2 is the centroid of the “Hot” membership function.

2. Determine the crisp value for fan speed based on the result of the calculation.

Let’s say the centroid of the “Normal” membership function is 40 and the centroid of the “Hot” membership function is 60 (these values represent the fan speeds). Using the formula above:

(0.8 * 40 + 0.3 * 60) / (0.8 + 0.3) ≈ 45.5

So, the crisp output for controlling the fan speed is approximately 45.5. In practical terms, this means that the fan should operate at a speed of around 45.5% of its maximum capacity to maintain the desired temperature.

Defuzzification takes fuzzy reasoning and turns it into a concrete action or decision. It allows us to effectively control systems based on fuzzy logic in real-world applications, ensuring that machines respond appropriately to complex, imprecise inputs.

Fuzzy logic allows computers to handle imprecise, human-like descriptions and make decisions based on the degree of truth rather than rigid binary choices. It’s a powerful tool that’s widely used in various fields, from temperature control systems to artificial intelligence. By incorporating fuzzy logic, we bring machines closer to understanding the subtleties of our language and the complexities of our world.

--

--

Brian Chan
Brian Chan

Written by Brian Chan

☕ Data Scientist | AI Engineer | Programmer | Trader, I write post from time to time