Automated construction of deep hierarchies using arbitrary clustering algorithms
Jeroen Tempels
23/12/2016
Automated construction of deep hierarchies
using arbitrary clustering algorithms [2]
Automated construction of deep hierarchies
using arbitrary clustering algorithms [2]
Automated construction of deep hierarchies
using arbitrary clustering algorithms [3] [4] [6] [7] [8]
Preprocessing
Sampling
Clustering
Feature extraction
Linear classification algorithm
Automated construction of deep hierarchies
using arbitrary clustering algorithms [3] [4] [6] [7] [8]
Preprocessing
Sampling
Clustering
Feature extraction
Linear classification algorithm
Automated construction of deep hierarchies
using arbitrary clustering algorithms [3] [4] [6] [7] [8]
Preprocessing
Sampling
Clustering
Feature extraction
Linear classification algorithm
Automated construction of deep hierarchies
using arbitrary clustering algorithms [3] [4] [6] [7] [8]
Preprocessing
Sampling
Clustering
Feature extraction
Linear classification algorithm
Automated construction of deep hierarchies
using arbitrary clustering algorithms [3] [4] [5] [6] [7] [8]
Preprocessing
Sampling
Clustering
Feature extraction
Linear classification algorithm
- Prototype model
- Connectivity model
- Graph model
Automated construction of deep hierarchies
using arbitrary clustering algorithms [3] [4] [6] [7] [8]
Preprocessing
Sampling
Clustering
Feature extraction
Linear classification algorithm
Automated construction of deep hierarchies
using arbitrary clustering algorithms [1] [3] [4]
Pipeline
- amount of layers
- clusters per layer
- ...
Results
- Working implementation
- Most time spent on implementation
- 73% (should get 78%)
Future Work
- Abstract k-means into any clustering
- Try cluster validity indices
- Test correlation parameters with CVI's
- Find optimal CVI (performance, speed,...)
- (Optional) Try other heuristics
References
