Chapter map

1. 1. Data
2. 2. Global centre
3. 3. Global distances
4. 4. Candidate k
5. 5. Assignments
6. 6. Within distances
7. 7. Centre separation
8. 8. Dispersion ratio
9. 9. Cluster penalty
10. 10. Separation factor
11. 11. Square
12. 12. Candidate comparison
13. 13. Outlier case
14. 14. Fuzzy version
15. 15. Initialization
16. 16. Claim

Visit the How Mathematics Works hub for the wider series and its connections to education, reasoning, science, computing, design and careers.

---

1. Data

A small example makes the mechanism visible. Use x=0,2,8,10.

A concrete situation

There are four one-dimensional observations.

What the mathematics does not license

State units. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

List points. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

2. Global centre

This is where modelling choices begin to affect evidence. Use v0=5.

A concrete situation

This is the one-cluster centre.

What the mathematics does not license

Centre definition must match distance choice. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Record centre. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

3. Global distances

The central mathematical idea is concrete. Compute 5,3,3,5.

A concrete situation

Absolute deviations sum to E1=16.

What the mathematics does not license

Do not use squared distances silently. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Show terms. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

4. Candidate k

Here the geometry becomes operational. Set k=2.

A concrete situation

Candidate centres are 1 and 9.

What the mathematics does not license

PBM compares across k. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

State k. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

5. Assignments

This step deserves a careful pause. Group {0,2} and {8,10}.

A concrete situation

Each point uses its cluster centre.

What the mathematics does not license

Fuzzy versions use membership weights. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Record groups. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

6. Within distances

The useful question is not only how to calculate, but what the calculation assumes. Compute 1,1,1,1.

A concrete situation

Ek=4.

What the mathematics does not license

Keep the same distance norm. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Show terms. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

7. Centre separation

A small example makes the mechanism visible. Compute |9-1|.

A concrete situation

Dk=8.

What the mathematics does not license

Use maximum pairwise centre separation. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Show pair. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

8. Dispersion ratio

This is where modelling choices begin to affect evidence. Compute E1/Ek.

A concrete situation

16/4=4.

What the mathematics does not license

Small Ek increases the score. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Show ratio. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

9. Cluster penalty

The central mathematical idea is concrete. Multiply by 1/k.

A concrete situation

(1/2)x4=2.

What the mathematics does not license

Penalty discourages excessive k. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Show step. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

10. Separation factor

Here the geometry becomes operational. Multiply 2 by 8.

A concrete situation

Product=16.

What the mathematics does not license

Outlying centres can inflate Dk. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Show step. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

11. Square

This step deserves a careful pause. Compute 16 squared.

A concrete situation

PBM=256.

What the mathematics does not license

Name the squared convention. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Report value. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

12. Candidate comparison

The useful question is not only how to calculate, but what the calculation assumes. Calculate other k values.

A concrete situation

Select a relative maximum.

What the mathematics does not license

One score has no standalone scale. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Build table. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

13. Outlier case

A small example makes the mechanism visible. Add a distant observation.

A concrete situation

E1 and Dk may grow.

What the mathematics does not license

Maximum separation can be fragile. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Run sensitivity. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

14. Fuzzy version

This is where modelling choices begin to affect evidence. Replace hard assignments with membership weights.

A concrete situation

Ek becomes membership weighted.

What the mathematics does not license

Name exponent and exact formula. Keeping this boundary clear makes the result easier to audit and transfer.

Try it, check it, transfer it

Report version. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

15. Initialization

The central mathematical idea is concrete. Repeat fits for each k.

A concrete situation

Ek and centres may vary.

What the mathematics does not license

A poor local optimum distorts PBM. A trustworthy workflow records the choice before seeing the final test result.

Try it, check it, transfer it

Use restarts. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

16. Claim

Here the geometry becomes operational. Name distance,k range,E1,Ek,Dk and version.

A concrete situation

Teaching PBM=256.

What the mathematics does not license

Do not call it natural truth. Evidence becomes stronger when the limitation is tested instead of hidden.

Try it, check it, transfer it

Bound conclusion. Then explain the result to someone who has not seen the formula. If the explanation needs a hidden assumption, add that assumption to the model record.

---

A seven-day practice lab

1. Day 1: define the decision, observation unit, X variables and Y or reference condition.
2. Day 2: reproduce the smallest worked example and check every dimension, unit and sign.
3. Day 3: build a transparent baseline and write what it assumes.
4. Day 4: fit the main method inside a training-only pipeline; record every parameter.
5. Day 5: challenge one assumption with a deliberately difficult dataset or operating regime.
6. Day 6: compare validation metrics with the practical workload created by errors.
7. Day 7: present one supported conclusion, one unresolved uncertainty and one safe next experiment.

Questions for a student, parent or mentor

• Which part of the model is a mathematical definition, and which part is a choice?
• What small calculation would catch a sign, scale or dimension error?
• What evidence comes from training, validation and the untouched test?
• What benign condition could look like a fault or prediction failure?
• How would the answer change for another response, asset, school project or time period?
• What should a human review before acting on the score?

A method-specific audit before deployment

The PBM index combines three ideas: overall dispersion around one global centre, within-cluster dispersion for a candidate partition, and the largest separation between candidate cluster centres. In a common crisp form, PBM(k)=[(1/k)(E1/Ek)Dk]^2. Larger values favour fewer compact clusters whose centres include a wide separation.

Take one-dimensional observations 0, 2, 8 and 10. Their global centre is 5, so E1=|0-5|+|2-5|+|8-5|+|10-5|=5+3+3+5=16. For k=2 with centres 1 and 9 and assignments {0,2} and {8,10}, Ek=1+1+1+1=4.

The largest centre separation is Dk=|9-1|=8. Therefore PBM(2)=[(1/2)(16/4)8]^2=[16]^2=256. The number is meaningful only relative to PBM values for other candidate partitions built from the same data, distance and conventions.

PBM can be influenced by outliers because E1 and maximum centre separation can grow. A single distant centre may inflate Dk, and different distance powers or fuzzy membership weights change the calculation. Report the exact version, candidate range, preprocessing, initialization and sensitivity rather than treating the largest score as unquestionable truth.

---

Frequently asked questions

What is the PBM index?

A squared product combining a cluster-count penalty, global-to-within dispersion ratio and maximum centre separation.

Is higher better?

Across comparable candidate partitions, the relative maximum is preferred under the named version.

What are E1 and Ek?

E1 is one-cluster dispersion around a global centre; Ek is within-cluster dispersion for k clusters.

What is Dk?

The maximum distance between candidate cluster centres.

Why can outliers matter?

They can increase global dispersion and maximum centre separation.

What must be reported?

Exact crisp or fuzzy formula, distance, centres, k range, E1, Ek, Dk, scaling, starts, score table and sensitivity checks.

---

Primary source and fact-check note

Malay K. Pakhira, Sanghamitra Bandyopadhyay and Ujjwal Maulik, “Validity Index for Crisp and Fuzzy Clusters,” Pattern Recognition 37(3), 487–501 (2004), DOI 10.1016/j.patcog.2003.06.005. Definitions and core claims in this article were checked against this primary publication. Calculations shown here were independently recomputed. Examples are educational illustrations; they are not reported industrial performance results.

Useful next reading

• Xie–Beni fuzzy validity index
• Fuzzy silhouette
• Consensus clustering
• Feature scaling and distance geometry
• How Mathematics Works: Statistics

Keep the options open. Mathematics education supports clearer questions, transferable problem-solving skills and informed choices across science, engineering, computing, AI, statistics, design and many other careers. It does not guarantee intelligence, income, admission or employment; those outcomes depend on far more than one subject or model.

Extended laboratory: fuzzy clustering evidence and responsible use

Data audit

Freeze observations, identities, units, exclusions, missing values, duplicates, transformations and seeds.

Membership audit

Verify row normalization, cluster count, fuzzifier, top memberships, hard-label convention and zero treatment.

Formula audit

Show logarithms, squared degrees, baseline transforms, weighted silhouettes, dispersions, centre gaps and denominators.

Convention audit

Name log base, 0 log 0 rule, alpha, distance norm, crisp or fuzzy PBM version and higher-or-lower direction.

Uncertainty audit

Repeat random starts, resamples, candidate counts, fuzzifiers, feature scales and plausible preprocessing choices.

Failure audit

Test uniform memberships, ties, singletons, tiny clusters, local optima, outliers and coincident or distant centres.

Meaning audit

Compare numerical crispness or geometry with domain evidence without assuming groups are natural, fair or useful.

Final claim

State exact uncertainty, crispness, weighted separation or dispersion evidence and stop before causal or guaranteed-usefulness claims.

Mechanism workshop: calculate, challenge and transfer

Workshop 1: Data

Begin with this fuzzy-clustering mechanism: Use x=0,2,8,10. Recreate the concrete case: There are four one-dimensional observations. Write observations, global centre, candidate centres, assignments or memberships, E1, Ek, k, Dk, product and squared index with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: State units. Complete the transfer step: List points. Change exactly one membership degree, cluster count, logarithm base, zero convention, fuzzifier, membership-gap exponent, hard label, singleton rule, feature scale, distance, centre, outlier, initialization or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep membership uncertainty, crispness, cohesion, separation, stability and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving uniform memberships, a top-membership tie, one outlier, a singleton, a poor local optimum, a distant centre or incompatible distance conventions. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 2: Global centre

Begin with this fuzzy-clustering mechanism: Use v0=5. Recreate the concrete case: This is the one-cluster centre. Write observations, global centre, candidate centres, assignments or memberships, E1, Ek, k, Dk, product and squared index with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Centre definition must match distance choice. Complete the transfer step: Record centre. Change exactly one membership degree, cluster count, logarithm base, zero convention, fuzzifier, membership-gap exponent, hard label, singleton rule, feature scale, distance, centre, outlier, initialization or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep membership uncertainty, crispness, cohesion, separation, stability and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving uniform memberships, a top-membership tie, one outlier, a singleton, a poor local optimum, a distant centre or incompatible distance conventions. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 3: Global distances

Begin with this fuzzy-clustering mechanism: Compute 5,3,3,5. Recreate the concrete case: Absolute deviations sum to E1=16. Write observations, global centre, candidate centres, assignments or memberships, E1, Ek, k, Dk, product and squared index with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Do not use squared distances silently. Complete the transfer step: Show terms. Change exactly one membership degree, cluster count, logarithm base, zero convention, fuzzifier, membership-gap exponent, hard label, singleton rule, feature scale, distance, centre, outlier, initialization or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep membership uncertainty, crispness, cohesion, separation, stability and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving uniform memberships, a top-membership tie, one outlier, a singleton, a poor local optimum, a distant centre or incompatible distance conventions. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 4: Candidate k

Begin with this fuzzy-clustering mechanism: Set k=2. Recreate the concrete case: Candidate centres are 1 and 9. Write observations, global centre, candidate centres, assignments or memberships, E1, Ek, k, Dk, product and squared index with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: PBM compares across k. Complete the transfer step: State k. Change exactly one membership degree, cluster count, logarithm base, zero convention, fuzzifier, membership-gap exponent, hard label, singleton rule, feature scale, distance, centre, outlier, initialization or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep membership uncertainty, crispness, cohesion, separation, stability and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving uniform memberships, a top-membership tie, one outlier, a singleton, a poor local optimum, a distant centre or incompatible distance conventions. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 5: Assignments

Begin with this fuzzy-clustering mechanism: Group {0,2} and {8,10}. Recreate the concrete case: Each point uses its cluster centre. Write observations, global centre, candidate centres, assignments or memberships, E1, Ek, k, Dk, product and squared index with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Fuzzy versions use membership weights. Complete the transfer step: Record groups. Change exactly one membership degree, cluster count, logarithm base, zero convention, fuzzifier, membership-gap exponent, hard label, singleton rule, feature scale, distance, centre, outlier, initialization or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep membership uncertainty, crispness, cohesion, separation, stability and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving uniform memberships, a top-membership tie, one outlier, a singleton, a poor local optimum, a distant centre or incompatible distance conventions. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 6: Within distances

Begin with this fuzzy-clustering mechanism: Compute 1,1,1,1. Recreate the concrete case: Ek=4. Write observations, global centre, candidate centres, assignments or memberships, E1, Ek, k, Dk, product and squared index with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Keep the same distance norm. Complete the transfer step: Show terms. Change exactly one membership degree, cluster count, logarithm base, zero convention, fuzzifier, membership-gap exponent, hard label, singleton rule, feature scale, distance, centre, outlier, initialization or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep membership uncertainty, crispness, cohesion, separation, stability and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving uniform memberships, a top-membership tie, one outlier, a singleton, a poor local optimum, a distant centre or incompatible distance conventions. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 7: Centre separation

Begin with this fuzzy-clustering mechanism: Compute |9-1|. Recreate the concrete case: Dk=8. Write observations, global centre, candidate centres, assignments or memberships, E1, Ek, k, Dk, product and squared index with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Use maximum pairwise centre separation. Complete the transfer step: Show pair. Change exactly one membership degree, cluster count, logarithm base, zero convention, fuzzifier, membership-gap exponent, hard label, singleton rule, feature scale, distance, centre, outlier, initialization or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep membership uncertainty, crispness, cohesion, separation, stability and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving uniform memberships, a top-membership tie, one outlier, a singleton, a poor local optimum, a distant centre or incompatible distance conventions. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 8: Dispersion ratio

Begin with this fuzzy-clustering mechanism: Compute E1/Ek. Recreate the concrete case: 16/4=4. Write observations, global centre, candidate centres, assignments or memberships, E1, Ek, k, Dk, product and squared index with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Small Ek increases the score. Complete the transfer step: Show ratio. Change exactly one membership degree, cluster count, logarithm base, zero convention, fuzzifier, membership-gap exponent, hard label, singleton rule, feature scale, distance, centre, outlier, initialization or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep membership uncertainty, crispness, cohesion, separation, stability and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving uniform memberships, a top-membership tie, one outlier, a singleton, a poor local optimum, a distant centre or incompatible distance conventions. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 9: Cluster penalty

Begin with this fuzzy-clustering mechanism: Multiply by 1/k. Recreate the concrete case: (1/2)x4=2. Write observations, global centre, candidate centres, assignments or memberships, E1, Ek, k, Dk, product and squared index with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Penalty discourages excessive k. Complete the transfer step: Show step. Change exactly one membership degree, cluster count, logarithm base, zero convention, fuzzifier, membership-gap exponent, hard label, singleton rule, feature scale, distance, centre, outlier, initialization or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep membership uncertainty, crispness, cohesion, separation, stability and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving uniform memberships, a top-membership tie, one outlier, a singleton, a poor local optimum, a distant centre or incompatible distance conventions. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 10: Separation factor

Begin with this fuzzy-clustering mechanism: Multiply 2 by 8. Recreate the concrete case: Product=16. Write observations, global centre, candidate centres, assignments or memberships, E1, Ek, k, Dk, product and squared index with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Outlying centres can inflate Dk. Complete the transfer step: Show step. Change exactly one membership degree, cluster count, logarithm base, zero convention, fuzzifier, membership-gap exponent, hard label, singleton rule, feature scale, distance, centre, outlier, initialization or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep membership uncertainty, crispness, cohesion, separation, stability and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving uniform memberships, a top-membership tie, one outlier, a singleton, a poor local optimum, a distant centre or incompatible distance conventions. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 11: Square

Begin with this fuzzy-clustering mechanism: Compute 16 squared. Recreate the concrete case: PBM=256. Write observations, global centre, candidate centres, assignments or memberships, E1, Ek, k, Dk, product and squared index with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Name the squared convention. Complete the transfer step: Report value. Change exactly one membership degree, cluster count, logarithm base, zero convention, fuzzifier, membership-gap exponent, hard label, singleton rule, feature scale, distance, centre, outlier, initialization or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep membership uncertainty, crispness, cohesion, separation, stability and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving uniform memberships, a top-membership tie, one outlier, a singleton, a poor local optimum, a distant centre or incompatible distance conventions. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 12: Candidate comparison

Begin with this fuzzy-clustering mechanism: Calculate other k values. Recreate the concrete case: Select a relative maximum. Write observations, global centre, candidate centres, assignments or memberships, E1, Ek, k, Dk, product and squared index with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: One score has no standalone scale. Complete the transfer step: Build table. Change exactly one membership degree, cluster count, logarithm base, zero convention, fuzzifier, membership-gap exponent, hard label, singleton rule, feature scale, distance, centre, outlier, initialization or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep membership uncertainty, crispness, cohesion, separation, stability and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving uniform memberships, a top-membership tie, one outlier, a singleton, a poor local optimum, a distant centre or incompatible distance conventions. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 13: Outlier case

Begin with this fuzzy-clustering mechanism: Add a distant observation. Recreate the concrete case: E1 and Dk may grow. Write observations, global centre, candidate centres, assignments or memberships, E1, Ek, k, Dk, product and squared index with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Maximum separation can be fragile. Complete the transfer step: Run sensitivity. Change exactly one membership degree, cluster count, logarithm base, zero convention, fuzzifier, membership-gap exponent, hard label, singleton rule, feature scale, distance, centre, outlier, initialization or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep membership uncertainty, crispness, cohesion, separation, stability and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving uniform memberships, a top-membership tie, one outlier, a singleton, a poor local optimum, a distant centre or incompatible distance conventions. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 14: Fuzzy version

Begin with this fuzzy-clustering mechanism: Replace hard assignments with membership weights. Recreate the concrete case: Ek becomes membership weighted. Write observations, global centre, candidate centres, assignments or memberships, E1, Ek, k, Dk, product and squared index with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Name exponent and exact formula. Complete the transfer step: Report version. Change exactly one membership degree, cluster count, logarithm base, zero convention, fuzzifier, membership-gap exponent, hard label, singleton rule, feature scale, distance, centre, outlier, initialization or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep membership uncertainty, crispness, cohesion, separation, stability and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving uniform memberships, a top-membership tie, one outlier, a singleton, a poor local optimum, a distant centre or incompatible distance conventions. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 15: Initialization

Begin with this fuzzy-clustering mechanism: Repeat fits for each k. Recreate the concrete case: Ek and centres may vary. Write observations, global centre, candidate centres, assignments or memberships, E1, Ek, k, Dk, product and squared index with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: A poor local optimum distorts PBM. Complete the transfer step: Use restarts. Change exactly one membership degree, cluster count, logarithm base, zero convention, fuzzifier, membership-gap exponent, hard label, singleton rule, feature scale, distance, centre, outlier, initialization or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep membership uncertainty, crispness, cohesion, separation, stability and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving uniform memberships, a top-membership tie, one outlier, a singleton, a poor local optimum, a distant centre or incompatible distance conventions. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.

Workshop 16: Claim

Begin with this fuzzy-clustering mechanism: Name distance,k range,E1,Ek,Dk and version. Recreate the concrete case: Teaching PBM=256. Write observations, global centre, candidate centres, assignments or memberships, E1, Ek, k, Dk, product and squared index with identifiers and assumptions. Calculate the smallest non-trivial example by hand and reproduce it independently. Challenge this limitation: Do not call it natural truth. Complete the transfer step: Bound conclusion. Change exactly one membership degree, cluster count, logarithm base, zero convention, fuzzifier, membership-gap exponent, hard label, singleton rule, feature scale, distance, centre, outlier, initialization or random seed. Explain which mathematical object changes, which score moves and which interpretation must weaken. Keep membership uncertainty, crispness, cohesion, separation, stability and practical usefulness on separate notebook lines. Explain once to a student and once to a parent. Add a counterexample involving uniform memberships, a top-membership tie, one outlier, a singleton, a poor local optimum, a distant centre or incompatible distance conventions. State the extra evidence needed to distinguish explanations. Finish with one supported sentence, one unsupported sentence and one useful next measurement.