Vector Representation of Graphs: Application to the Classification of Symbols and Letters
read more
Citations
Fuzzy multilevel graph embedding
New binary linear programming formulation to compute the graph edit distance
A Comparison of Explicit and Implicit Graph Embedding Methods for Pattern Recognition
Contribution à l'analyse complexe de documents anciens, application aux lettrines
Graph Kernels in chemoinformatics
References
Computers and Intractability: A Guide to the Theory of NP-Completeness
Multi-Interval Discretization of Continuous-Valued Attributes for Classification Learning
An Algorithm for Subgraph Isomorphism
Thirty years of graph matching in pattern recognition
Related Papers (5)
Frequently Asked Questions (9)
Q2. What are the future works mentioned in the paper "Vector representation of graphs: application to the classification of symbols and letters" ?
Their future work will consist on showing the gain of time in this case of use.
Q3. How are the distortions applied to graphs?
In order to test classifiers under different conditions, distortions are applied on the prototype graphs with three different levels of strength, low, medium and high.
Q4. How can the authors build a graph with rank n?
At each iteration, it is possible to build a pattern of rank n by adding an edge to a pattern of rank n− 1 with the ability to add a vertex if needed.
Q5. How many classes are distorted in the graphs?
For an adequately sized set, all graphs are distorted nine times to obtain a data set containing 1,100 graphs uniformely distributed over the 22 classes.
Q6. What is the choice for the number of classes?
The optimal size of the lexicon depends onthe size (mean of nodes and edges) of the graphs and the best choice for the number of classes depends on attributes.
Q7. What is the effect of the distortion of the letters on the graph?
As the topology of the graph takes an important an important place in their representation, the distortion of the letters impacts on the results.
Q8. What is the classification result of a k-nearest neighbor classifier?
For each dataset, the classification result of a k-nearest neighbor classifier (k-NN) used with graph edit distance which is the reference, and with the vector description.
Q9. How are graphs extracted from the resulting denoised images?
graphs are extracted from the resulting denoised images by tracing the lines from end to end and detecting intersections as well as corners.