---
title: "Hopfield vs Kohonen network"
subject: "Artificial Intelligence"
chapter: 6
key: "semester_6/artificial_intelligence:hopfield-vs-kohonen"
source: https://www.bctengineer.com/q/semester_6/artificial_intelligence/hopfield-vs-kohonen
site: BCT Engineering
---

# Hopfield vs Kohonen network

Artificial Intelligence · Chapter 6: Machine Learning

- Asked in: 2071 Magh, 3 marks
- Answer checked by Claude Opus 5.5 (high effort) on 2026-10-14
- Page: https://www.bctengineer.com/q/semester_6/artificial_intelligence/hopfield-vs-kohonen
- Whole chapter: https://www.bctengineer.com/md/semester_6/artificial_intelligence/chapter-6

## Question

Compare Hopfield versus Kohonen network.

## Short answer

Hopfield network is a recurrent auto-associative memory with symmetric weights used for pattern completion. Kohonen SOM is unsupervised competitive learning for clustering and topology-preserving mapping. Hopfield iterates to stable states; SOM finds best matching unit.

## Answer

| Point | Hopfield network | Kohonen network (SOM) |
|---|---|---|
| Type of learning | Hebbian-type, effectively supervised storage of patterns | Unsupervised, competitive learning |
| Architecture | Single layer, fully connected, recurrent | Input layer and a 1-D/2-D output (map) layer, feed-forward |
| Connections | Symmetric weights, no self-loops | Each input linked to every output node; lateral competition |
| Purpose | Auto-associative memory, pattern completion, optimisation | Clustering, topology-preserving mapping, dimensionality reduction |
| Weight setting | Computed once: W = Σ pᵀp (diagonal 0) | Updated iteratively: w ← w + α(x − w) for winner and neighbours |
| Neuron output | Bipolar/binary, threshold function | Winner-takes-all (closest weight vector wins) |
| Recall | Iterates until a stable state (energy minimum) | Finds best matching unit for an input |
| Capacity/limit | About 0.15 N patterns; spurious states | Needs choice of map size, α and neighbourhood |

## Check yourself

The answer's key points as multiple-choice questions, with the correct option and why.

1. Which statement is correct?
   A. Hopfield uses competitive learning; Kohonen uses Hebbian storage
   B. Kohonen SOM uses unsupervised competitive learning; Hopfield uses Hebbian-type storage of patterns
   C. Both use back-propagation
   D. Both are feed-forward
   Answer: B. SOM neurons compete; Hopfield weights come from pattern correlations. (Key point: Learning type)

2. How are weights set in each?
   A. Both are computed once
   B. Hopfield iterates; Kohonen is computed once
   C. Hopfield computes W once as sum pT p; Kohonen updates iteratively w <- w + alpha(x - w)
   D. Both use gradient descent on error
   Answer: C. Kohonen moves the winner and neighbours toward the input. (Key point: Weight setting)

3. Which pairing of purpose is right?
   A. Hopfield: clustering; Kohonen: pattern completion
   B. Both: regression
   C. Both: logic gates only
   D. Hopfield: auto-associative memory; Kohonen: clustering and topology-preserving mapping
   Answer: D. Hopfield completes patterns; SOM maps and clusters. (Key point: Purpose)

## Related in Artificial Intelligence

- [Hopfield network and its steps](https://www.bctengineer.com/question.md?key=semester_6%2Fartificial_intelligence%3Ahopfield-network-steps) — Artificial Intelligence, chapter 6
- [Neural network; supervised vs unsupervised](https://www.bctengineer.com/question.md?key=semester_6%2Fartificial_intelligence%3Ann-supervised-unsupervised) — Artificial Intelligence, chapter 6

## Related in other subjects

- [Context switching; distributed vs network OS](https://www.bctengineer.com/question.md?key=semester_7%2Fdistributed_system%3Acontext-switch-dos) — Distributed System, chapter 3
- [Network OS vs distributed OS](https://www.bctengineer.com/question.md?key=semester_7%2Fdistributed_system%3Anos-vs-dos-diff) — Distributed System, chapter 3
- [Crossbar switch vs multistage network](https://www.bctengineer.com/question.md?key=semester_5%2Fcomputer_organization_and_architecture%3Acrossbar-vs-multistage) — Computer Organization and Architecture, chapter 8
- [Compare MOCB/BOCB and ABCB/puffer SF6](https://www.bctengineer.com/question.md?key=semester_6%2Fswitchgear_and_protection%3Acompare-mocb-bocb-abcb-sf6) — Switchgear and Protection, chapter 6