OpenI
/
graphkit-learn
Not watched
1
0
Fork 0
Code
Releases 2 Wiki Activity
Issues 0 Pull Requests 0 Datasets Model Cloudbrain
You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
76 Commits
4 Branches
345 MB
96 Download
Tree: 1566139883
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '1566139883'
${ noResults }
graphkit-learn/notebooks/get_dataset_attributes.ipynb

get_dataset_attributes.ipynb 23 kB
Raw Normal View History

Plus de choses sont changées...
7 years ago
​
​
​
​
​
​
​
​
​
​
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
​
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
​
1. [IMPORTANT] correct an error in dataset split section in function cross_validation_pre_computed. DON NOT use previous versions. They are all wrong. 2. save gram matrices and relative data when using function cross_validation_pre_computed, before cross validation step, in case that something goes wrong with CV. Parameter read_gm_from_file can be used to choose whether to read gram matrices from file. 3. add some test code to check if a gram matrix is symmetric and positive semi-definite.
6 years ago
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
1. [IMPORTANT] correct an error in dataset split section in function cross_validation_pre_computed. DON NOT use previous versions. They are all wrong. 2. save gram matrices and relative data when using function cross_validation_pre_computed, before cross validation step, in case that something goes wrong with CV. Parameter read_gm_from_file can be used to choose whether to read gram matrices from file. 3. add some test code to check if a gram matrix is symmetric and positive semi-definite.
6 years ago
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
1. [IMPORTANT] correct an error in dataset split section in function cross_validation_pre_computed. DON NOT use previous versions. They are all wrong. 2. save gram matrices and relative data when using function cross_validation_pre_computed, before cross validation step, in case that something goes wrong with CV. Parameter read_gm_from_file can be used to choose whether to read gram matrices from file. 3. add some test code to check if a gram matrix is symmetric and positive semi-definite.
6 years ago
​
​
​
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
​
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
​
Plus de choses sont changées...
7 years ago
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
Plus de choses sont changées...
7 years ago
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
Plus de choses sont changées...
7 years ago
​
1. [IMPORTANT] correct an error in dataset split section in function cross_validation_pre_computed. DON NOT use previous versions. They are all wrong. 2. save gram matrices and relative data when using function cross_validation_pre_computed, before cross validation step, in case that something goes wrong with CV. Parameter read_gm_from_file can be used to choose whether to read gram matrices from file. 3. add some test code to check if a gram matrix is symmetric and positive semi-definite.
6 years ago
​
Plus de choses sont changées...
7 years ago
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
​
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
​
Plus de choses sont changées...
7 years ago
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
Plus de choses sont changées...
7 years ago
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
Plus de choses sont changées...
7 years ago
​
​
1. [IMPORTANT] correct an error in dataset split section in function cross_validation_pre_computed. DON NOT use previous versions. They are all wrong. 2. save gram matrices and relative data when using function cross_validation_pre_computed, before cross validation step, in case that something goes wrong with CV. Parameter read_gm_from_file can be used to choose whether to read gram matrices from file. 3. add some test code to check if a gram matrix is symmetric and positive semi-definite.
6 years ago
Plus de choses sont changées...
7 years ago
​
​
​
​
​
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
Plus de choses sont changées...
7 years ago
​
​
​
​
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
Plus de choses sont changées...
7 years ago
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
Plus de choses sont changées...
7 years ago
​
1. [IMPORTANT] correct an error in dataset split section in function cross_validation_pre_computed. DON NOT use previous versions. They are all wrong. 2. save gram matrices and relative data when using function cross_validation_pre_computed, before cross validation step, in case that something goes wrong with CV. Parameter read_gm_from_file can be used to choose whether to read gram matrices from file. 3. add some test code to check if a gram matrix is symmetric and positive semi-definite.
6 years ago
​
Plus de choses sont changées...
7 years ago
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
Plus de choses sont changées...
7 years ago
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
Plus de choses sont changées...
7 years ago
​
1. [IMPORTANT] correct an error in dataset split section in function cross_validation_pre_computed. DON NOT use previous versions. They are all wrong. 2. save gram matrices and relative data when using function cross_validation_pre_computed, before cross validation step, in case that something goes wrong with CV. Parameter read_gm_from_file can be used to choose whether to read gram matrices from file. 3. add some test code to check if a gram matrix is symmetric and positive semi-definite.
6 years ago
​
Plus de choses sont changées...
7 years ago
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
Plus de choses sont changées...
7 years ago
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
Plus de choses sont changées...
7 years ago
​
​
1. [IMPORTANT] correct an error in dataset split section in function cross_validation_pre_computed. DON NOT use previous versions. They are all wrong. 2. save gram matrices and relative data when using function cross_validation_pre_computed, before cross validation step, in case that something goes wrong with CV. Parameter read_gm_from_file can be used to choose whether to read gram matrices from file. 3. add some test code to check if a gram matrix is symmetric and positive semi-definite.
6 years ago
Plus de choses sont changées...
7 years ago
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
Plus de choses sont changées...
7 years ago
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
Plus de choses sont changées...
7 years ago
​
​
1. [IMPORTANT] correct an error in dataset split section in function cross_validation_pre_computed. DON NOT use previous versions. They are all wrong. 2. save gram matrices and relative data when using function cross_validation_pre_computed, before cross validation step, in case that something goes wrong with CV. Parameter read_gm_from_file can be used to choose whether to read gram matrices from file. 3. add some test code to check if a gram matrix is symmetric and positive semi-definite.
6 years ago
Plus de choses sont changées...
7 years ago
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
​
Plus de choses sont changées...
7 years ago
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
Plus de choses sont changées...
7 years ago
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
Plus de choses sont changées...
7 years ago
​
​
1. [IMPORTANT] correct an error in dataset split section in function cross_validation_pre_computed. DON NOT use previous versions. They are all wrong. 2. save gram matrices and relative data when using function cross_validation_pre_computed, before cross validation step, in case that something goes wrong with CV. Parameter read_gm_from_file can be used to choose whether to read gram matrices from file. 3. add some test code to check if a gram matrix is symmetric and positive semi-definite.
6 years ago
Plus de choses sont changées...
7 years ago
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
Plus de choses sont changées...
7 years ago
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
Plus de choses sont changées...
7 years ago
​
​
1. [IMPORTANT] correct an error in dataset split section in function cross_validation_pre_computed. DON NOT use previous versions. They are all wrong. 2. save gram matrices and relative data when using function cross_validation_pre_computed, before cross validation step, in case that something goes wrong with CV. Parameter read_gm_from_file can be used to choose whether to read gram matrices from file. 3. add some test code to check if a gram matrix is symmetric and positive semi-definite.
6 years ago
Plus de choses sont changées...
7 years ago
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
Plus de choses sont changées...
7 years ago
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
Plus de choses sont changées...
7 years ago
​
​
1. [IMPORTANT] correct an error in dataset split section in function cross_validation_pre_computed. DON NOT use previous versions. They are all wrong. 2. save gram matrices and relative data when using function cross_validation_pre_computed, before cross validation step, in case that something goes wrong with CV. Parameter read_gm_from_file can be used to choose whether to read gram matrices from file. 3. add some test code to check if a gram matrix is symmetric and positive semi-definite.
6 years ago
Plus de choses sont changées...
7 years ago
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
Plus de choses sont changées...
7 years ago
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
Plus de choses sont changées...
7 years ago
​
​
1. [IMPORTANT] correct an error in dataset split section in function cross_validation_pre_computed. DON NOT use previous versions. They are all wrong. 2. save gram matrices and relative data when using function cross_validation_pre_computed, before cross validation step, in case that something goes wrong with CV. Parameter read_gm_from_file can be used to choose whether to read gram matrices from file. 3. add some test code to check if a gram matrix is symmetric and positive semi-definite.
6 years ago
Plus de choses sont changées...
7 years ago
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
Plus de choses sont changées...
7 years ago
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
Plus de choses sont changées...
7 years ago
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
Plus de choses sont changées...
7 years ago
​
1. [IMPORTANT] correct an error in dataset split section in function cross_validation_pre_computed. DON NOT use previous versions. They are all wrong. 2. save gram matrices and relative data when using function cross_validation_pre_computed, before cross validation step, in case that something goes wrong with CV. Parameter read_gm_from_file can be used to choose whether to read gram matrices from file. 3. add some test code to check if a gram matrix is symmetric and positive semi-definite.
6 years ago
​
Plus de choses sont changées...
7 years ago
1. [IMPORTANT] correct an error in dataset split section in function cross_validation_pre_computed. DON NOT use previous versions. They are all wrong. 2. save gram matrices and relative data when using function cross_validation_pre_computed, before cross validation step, in case that something goes wrong with CV. Parameter read_gm_from_file can be used to choose whether to read gram matrices from file. 3. add some test code to check if a gram matrix is symmetric and positive semi-definite.
6 years ago
Plus de choses sont changées...
7 years ago
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
Plus de choses sont changées...
7 years ago
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
Plus de choses sont changées...
7 years ago
​
1. [IMPORTANT] correct an error in dataset split section in function cross_validation_pre_computed. DON NOT use previous versions. They are all wrong. 2. save gram matrices and relative data when using function cross_validation_pre_computed, before cross validation step, in case that something goes wrong with CV. Parameter read_gm_from_file can be used to choose whether to read gram matrices from file. 3. add some test code to check if a gram matrix is symmetric and positive semi-definite.
6 years ago
​
Plus de choses sont changées...
7 years ago
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
​
​
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
​
​
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
Plus de choses sont changées...
7 years ago
​
​
​
​
​
​
​
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
​
​
​
​
​
1. [IMPORTANT] correct an error in dataset split section in function cross_validation_pre_computed. DON NOT use previous versions. They are all wrong. 2. save gram matrices and relative data when using function cross_validation_pre_computed, before cross validation step, in case that something goes wrong with CV. Parameter read_gm_from_file can be used to choose whether to read gram matrices from file. 3. add some test code to check if a gram matrix is symmetric and positive semi-definite.
6 years ago
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
Plus de choses sont changées...
7 years ago
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
​
Plus de choses sont changées...
7 years ago
​
​
​
​
​
​
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
Plus de choses sont changées...
7 years ago
1. fix bug when computing kernels between non-symbolic node attributes of the spkernel and the ssp kernel, add a gaussian kernel implementation. 2. update pygraph.utils.graphdatasets.get_dataset_attributes function, so that if a dataset has missing attributes it can still read the dimension of attributes.
6 years ago
Plus de choses sont changées...
7 years ago
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
* use global variable to tackle big read-only variables when they are used in subprocesses. * rewrite the implementation of the marginalized kernel. * implement four computing methods of the generalized random walk kernel. * in the path kernel up to length h, use trie to save all paths, saving tremendous memories; use the Deep-first search to get paths from graphs. * in model_selection_for_precomputed_kernel method, complete the part to do cross validation when Gram matrices are read from file. * in get_dataset_attributes methods, correct three sub-methods about getting node degrees, add sub-methods to get fill factors of graphs. * change default chunksize of pool.imap_unordered parallel method to 100. * remove try... except blocks in case they hidden bugs.
6 years ago
Plus de choses sont changées...
7 years ago
​
​
​
​
 123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654 
  1. {

  2.  "cells": [

  3.   {

  4.    "cell_type": "code",

  5.    "execution_count": 1,

  6.    "metadata": {

  7.     "scrolled": false

  8.    },

  9.    "outputs": [

  10.     {

  11.      "name": "stdout",

  12.      "output_type": "stream",

  13.      "text": [

  14.       "\n",

  15.       "Acyclic:\n",

  16.       "substructures : {'non linear', 'linear'}\n",

  17.       "node_labeled : True\n",

  18.       "edge_labeled : False\n",

  19.       "is_directed : False\n",

  20.       "dataset_size : 183\n",

  21.       "ave_node_num : 8.153005464480874\n",

  22.       "min_node_num : 3\n",

  23.       "max_node_num : 11\n",

  24.       "ave_edge_num : 7.1530054644808745\n",

  25.       "min_edge_num : 2\n",

  26.       "max_edge_num : 10\n",

  27.       "ave_node_degree : 1.737561012151176\n",

  28.       "min_node_degree : 1.3333333333333333\n",

  29.       "max_node_degree : 1.8181818181818181\n",

  30.       "ave_fill_factor : 0.11241161841596808\n",

  31.       "min_fill_factor : 0.08264462809917356\n",

  32.       "max_fill_factor : 0.2222222222222222\n",

  33.       "node_label_num : 3\n",

  34.       "edge_label_num : 1\n",

  35.       "node_attr_dim : 0\n",

  36.       "edge_attr_dim : 0\n",

  37.       "class_number : 148\n",

  38.       "\n",

  39.       "\n",

  40.       "Alkane:\n",

  41.       "substructures : {'non linear', 'linear'}\n",

  42.       "node_labeled : True\n",

  43.       "edge_labeled : False\n",

  44.       "is_directed : False\n",

  45.       "dataset_size : 150\n",

  46.       "ave_node_num : 8.873333333333333\n",

  47.       "min_node_num : 1\n",

  48.       "max_node_num : 10\n",

  49.       "ave_edge_num : 7.873333333333333\n",

  50.       "min_edge_num : 0\n",

  51.       "max_edge_num : 9\n",

  52.       "ave_node_degree : 1.7507830687830694\n",

  53.       "min_node_degree : 0.0\n",

  54.       "max_node_degree : 1.8\n",

  55.       "ave_fill_factor : 0.10199498404299989\n",

  56.       "min_fill_factor : 0.0\n",

  57.       "max_fill_factor : 0.25\n",

  58.       "node_label_num : 2\n",

  59.       "edge_label_num : 1\n",

  60.       "node_attr_dim : 0\n",

  61.       "edge_attr_dim : 0\n",

  62.       "class_number : 123\n",

  63.       "\n",

  64.       "\n",

  65.       "MAO:\n",

  66.       "substructures : {'non linear', 'linear'}\n",

  67.       "node_labeled : True\n",

  68.       "edge_labeled : True\n",

  69.       "is_directed : False\n",

  70.       "dataset_size : 68\n",

  71.       "ave_node_num : 18.38235294117647\n",

  72.       "min_node_num : 11\n",

  73.       "max_node_num : 27\n",

  74.       "ave_edge_num : 19.63235294117647\n",

  75.       "min_edge_num : 12\n",

  76.       "max_edge_num : 29\n",

  77.       "ave_node_degree : 2.1347114940751464\n",

  78.       "min_node_degree : 2.090909090909091\n",

  79.       "max_node_degree : 2.2\n",

  80.       "ave_fill_factor : 0.060638921710159575\n",

  81.       "min_fill_factor : 0.039780521262002745\n",

  82.       "max_fill_factor : 0.09917355371900827\n",

  83.       "node_label_num : 3\n",

  84.       "edge_label_num : 4\n",

  85.       "node_attr_dim : 0\n",

  86.       "edge_attr_dim : 0\n",

  87.       "class_number : 2\n",

  88.       "\n",

  89.       "\n",

  90.       "PAH:\n",

  91.       "substructures : {'non linear', 'linear'}\n",

  92.       "node_labeled : False\n",

  93.       "edge_labeled : False\n",

  94.       "is_directed : False\n",

  95.       "dataset_size : 94\n",

  96.       "ave_node_num : 20.70212765957447\n",

  97.       "min_node_num : 10\n",

  98.       "max_node_num : 28\n",

  99.       "ave_edge_num : 24.425531914893618\n",

  100.       "min_edge_num : 11\n",

  101.       "max_edge_num : 34\n",

  102.       "ave_node_degree : 2.3550919704450077\n",

  103.       "min_node_degree : 2.2\n",

  104.       "max_node_degree : 2.5\n",

  105.       "ave_fill_factor : 0.05799294134806485\n",

  106.       "min_fill_factor : 0.04336734693877551\n",

  107.       "max_fill_factor : 0.11\n",

  108.       "node_label_num : 1\n",

  109.       "edge_label_num : 1\n",

  110.       "node_attr_dim : 0\n",

  111.       "edge_attr_dim : 0\n",

  112.       "class_number : 2\n",

  113.       "\n",

  114.       "\n",

  115.       "MUTAG:\n",

  116.       "substructures : {'non linear', 'linear'}\n",

  117.       "node_labeled : True\n",

  118.       "edge_labeled : True\n",

  119.       "is_directed : False\n",

  120.       "dataset_size : 188\n",

  121.       "ave_node_num : 17.930851063829788\n",

  122.       "min_node_num : 10\n",

  123.       "max_node_num : 28\n",

  124.       "ave_edge_num : 19.79255319148936\n",

  125.       "min_edge_num : 10\n",

  126.       "max_edge_num : 33\n",

  127.       "ave_node_degree : 2.1887720785524962\n",

  128.       "min_node_degree : 2.0\n",

  129.       "max_node_degree : 2.4444444444444446\n",

  130.       "ave_fill_factor : 0.06480462822996713\n",

  131.       "min_fill_factor : 0.039540816326530615\n",

  132.       "max_fill_factor : 0.1\n",

  133.       "node_label_num : 7\n",

  134.       "edge_label_num : 11\n",

  135.       "node_attr_dim : 0\n",

  136.       "edge_attr_dim : 0\n",

  137.       "class_number : 2\n",

  138.       "\n",

  139.       "\n",

  140.       "Letter-med:\n",

  141.       "substructures : {'non linear', 'linear'}\n",

  142.       "node_labeled : False\n",

  143.       "edge_labeled : False\n",

  144.       "is_directed : False\n",

  145.       "dataset_size : 2250\n",

  146.       "ave_node_num : 4.674666666666667\n",

  147.       "min_node_num : 1\n",

  148.       "max_node_num : 9\n",

  149.       "ave_edge_num : 3.2057777777777776\n",

  150.       "min_edge_num : 0\n",

  151.       "max_edge_num : 7\n",

  152.       "ave_node_degree : 1.35270582010582\n",

  153.       "min_node_degree : 0.0\n",

  154.       "max_node_degree : 2.4\n",

  155.       "ave_fill_factor : 0.15517701625094482\n",

  156.       "min_fill_factor : 0.0\n",

  157.       "max_fill_factor : 0.3333333333333333\n",

  158.       "node_label_num : 0\n",

  159.       "edge_label_num : 0\n",

  160.       "node_attr_dim : 2\n",

  161.       "edge_attr_dim : 0\n",

  162.       "class_number : 15\n",

  163.       "\n",

  164.       "\n",

  165.       "ENZYMES:\n",

  166.       "substructures : {'non linear', 'linear'}\n",

  167.       "node_labeled : True\n",

  168.       "edge_labeled : False\n",

  169.       "is_directed : False\n",

  170.       "dataset_size : 600\n",

  171.       "ave_node_num : 32.63333333333333\n",

  172.       "min_node_num : 2\n",

  173.       "max_node_num : 126\n",

  174.       "ave_edge_num : 62.13666666666666\n",

  175.       "min_edge_num : 1\n",

  176.       "max_edge_num : 149\n",

  177.       "ave_node_degree : 3.862625314410413\n",

  178.       "min_node_degree : 0.32\n",

  179.       "max_node_degree : 5.230769230769231\n",

  180.       "ave_fill_factor : 0.07509817146721588\n",

  181.       "min_fill_factor : 0.0016\n",

  182.       "max_fill_factor : 0.375\n",

  183.       "node_label_num : 3\n",

  184.       "edge_label_num : 0\n",

  185.       "node_attr_dim : 18\n",

  186.       "edge_attr_dim : 0\n",

  187.       "class_number : 6\n",

  188.       "\n",

  189.       "\n",

  190.       "Mutagenicity:\n",

  191.       "substructures : {'non linear', 'linear'}\n",

  192.       "node_labeled : True\n",

  193.       "edge_labeled : True\n",

  194.       "is_directed : False\n",

  195.       "dataset_size : 4337\n",

  196.       "ave_node_num : 30.317731150564907\n",

  197.       "min_node_num : 4\n",

  198.       "max_node_num : 417\n",

  199.       "ave_edge_num : 30.76942587041734\n",

  200.       "min_edge_num : 3\n",

  201.       "max_edge_num : 112\n",

  202.       "ave_node_degree : 2.0379886162441148\n",

  203.       "min_node_degree : 0.47961630695443647\n",

  204.       "max_node_degree : 2.3703703703703702\n",

  205.       "ave_fill_factor : 0.0431047931997047\n",

  206.       "min_fill_factor : 0.0005750795047415305\n",

  207.       "max_fill_factor : 0.1875\n",

  208.       "node_label_num : 14\n",

  209.       "edge_label_num : 3\n",

  210.       "node_attr_dim : 0\n",

  211.       "edge_attr_dim : 0\n",

  212.       "class_number : 2\n",

  213.       "\n",

  214.       "\n",

  215.       "D&D:\n",

  216.       "substructures : {'non linear', 'linear'}\n",

  217.       "node_labeled : True\n",

  218.       "edge_labeled : False\n",

  219.       "is_directed : False\n",

  220.       "dataset_size : 1178\n",

  221.       "ave_node_num : 284.3166383701188\n",

  222.       "min_node_num : 30\n",

  223.       "max_node_num : 5748\n",

  224.       "ave_edge_num : 715.6587436332767\n",

  225.       "min_edge_num : 63\n",

  226.       "max_edge_num : 14267\n",

  227.       "ave_node_degree : 4.979061662020889\n",

  228.       "min_node_degree : 3.6116504854368934\n",

  229.       "max_node_degree : 8.933333333333334\n",

  230.       "ave_fill_factor : 0.013790239865199101\n",

  231.       "min_fill_factor : 0.0004318164098347239\n",

  232.       "max_fill_factor : 0.09666666666666666\n",

  233.       "node_label_num : 82\n",

  234.       "edge_label_num : 0\n",

  235.       "node_attr_dim : 0\n",

  236.       "edge_attr_dim : 0\n",

  237.       "class_number : 2\n",

  238.       "\n",

  239.       "\n",

  240.       "AIDS:\n",

  241.       "substructures : {'non linear', 'linear'}\n",

  242.       "node_labeled : True\n",

  243.       "edge_labeled : True\n",

  244.       "is_directed : False\n",

  245.       "dataset_size : 2000\n",

  246.       "ave_node_num : 15.6925\n",

  247.       "min_node_num : 2\n",

  248.       "max_node_num : 95\n",

  249.       "ave_edge_num : 16.195\n",

  250.       "min_edge_num : 1\n",

  251.       "max_edge_num : 103\n",

  252.       "ave_node_degree : 2.012865369646626\n",

  253.       "min_node_degree : 0.6\n",

  254.       "max_node_degree : 2.8333333333333335\n",

  255.       "ave_fill_factor : 0.08679744688995196\n",

  256.       "min_fill_factor : 0.011412742382271468\n",

  257.       "max_fill_factor : 0.25\n",

  258.       "node_label_num : 38\n",

  259.       "edge_label_num : 3\n",

  260.       "node_attr_dim : 4\n",

  261.       "edge_attr_dim : 0\n",

  262.       "class_number : 2\n",

  263.       "\n",

  264.       "\n",

  265.       "FIRSTMM_DB:\n",

  266.       "substructures : {'non linear'}\n",

  267.       "node_labeled : True\n",

  268.       "edge_labeled : False\n",

  269.       "is_directed : False\n",

  270.       "dataset_size : 41\n",

  271.       "ave_node_num : 1377.2682926829268\n",

  272.       "min_node_num : 134\n",

  273.       "max_node_num : 5037\n",

  274.       "ave_edge_num : 3074.0975609756097\n",

  275.       "min_edge_num : 320\n",

  276.       "max_edge_num : 10888\n",

  277.       "ave_node_degree : 4.503061007447199\n",

  278.       "min_node_degree : 4.191919191919192\n",

  279.       "max_node_degree : 4.776119402985074\n",

  280.       "ave_fill_factor : 0.003689884678097613\n",

  281.       "min_fill_factor : 0.00042914515176536197\n",

  282.       "max_fill_factor : 0.017821341055914458\n",

  283.       "node_label_num : 5\n",

  284.       "edge_label_num : 0\n",

  285.       "node_attr_dim : 1\n",

  286.       "edge_attr_dim : 2\n",

  287.       "class_number : 11\n",

  288.       "\n",

  289.       "\n",

  290.       "MSRC9:\n",

  291.       "substructures : {'non linear', 'linear'}\n",

  292.       "node_labeled : True\n",

  293.       "edge_labeled : False\n",

  294.       "is_directed : False\n",

  295.       "dataset_size : 221\n",

  296.       "ave_node_num : 40.57918552036199\n",

  297.       "min_node_num : 25\n",

  298.       "max_node_num : 55\n",

  299.       "ave_edge_num : 97.9366515837104\n",

  300.       "min_edge_num : 53\n",

  301.       "max_edge_num : 145\n",

  302.       "ave_node_degree : 4.8153400199203436\n",

  303.       "min_node_degree : 4.176470588235294\n",

  304.       "max_node_degree : 5.576923076923077\n",

  305.       "ave_fill_factor : 0.06021937645679636\n",

  306.       "min_fill_factor : 0.04521181915272339\n",

  307.       "max_fill_factor : 0.0848\n",

  308.       "node_label_num : 10\n",

  309.       "edge_label_num : 0\n",

  310.       "node_attr_dim : 0\n",

  311.       "edge_attr_dim : 0\n",

  312.       "class_number : 8\n",

  313.       "\n",

  314.       "\n",

  315.       "MSRC21:\n",

  316.       "substructures : {'non linear', 'linear'}\n",

  317.       "node_labeled : True\n",

  318.       "edge_labeled : False\n",

  319.       "is_directed : False\n",

  320.       "dataset_size : 563\n",

  321.       "ave_node_num : 77.52042628774423\n",

  322.       "min_node_num : 51\n",

  323.       "max_node_num : 141\n",

  324.       "ave_edge_num : 198.32326820603907\n",

  325.       "min_edge_num : 121\n",

  326.       "max_edge_num : 405\n",

  327.       "ave_node_degree : 5.102391320310953\n",

  328.       "min_node_degree : 4.04\n",

  329.       "max_node_degree : 6.6\n",

  330.       "ave_fill_factor : 0.03357132864022473\n",

  331.       "min_fill_factor : 0.01873405612244898\n",

  332.       "max_fill_factor : 0.04652056901191849\n",

  333.       "node_label_num : 22\n",

  334.       "edge_label_num : 0\n",

  335.       "node_attr_dim : 0\n",

  336.       "edge_attr_dim : 0\n",

  337.       "class_number : 20\n",

  338.       "\n",

  339.       "\n",

  340.       "SYNTHETIC:\n",

  341.       "substructures : {'non linear', 'linear'}\n",

  342.       "node_labeled : True\n",

  343.       "edge_labeled : False\n",

  344.       "is_directed : False\n",

  345.       "dataset_size : 300\n",

  346.       "ave_node_num : 100.0\n",

  347.       "min_node_num : 100\n",

  348.       "max_node_num : 100\n",

  349.       "ave_edge_num : 196.0\n",

  350.       "min_edge_num : 196\n",

  351.       "max_edge_num : 196\n",

  352.       "ave_node_degree : 3.9200000000000017\n",

  353.       "min_node_degree : 3.92\n",

  354.       "max_node_degree : 3.92\n",

  355.       "ave_fill_factor : 0.019600000000000003\n",

  356.       "min_fill_factor : 0.0196\n",

  357.       "max_fill_factor : 0.0196\n",

  358.       "node_label_num : 8\n",

  359.       "edge_label_num : 0\n",

  360.       "node_attr_dim : 1\n",

  361.       "edge_attr_dim : 0\n",

  362.       "class_number : 2\n",

  363.       "\n",

  364.       "\n",

  365.       "BZR:\n",

  366.       "substructures : {'non linear', 'linear'}\n",

  367.       "node_labeled : True\n",

  368.       "edge_labeled : False\n",

  369.       "is_directed : False\n",

  370.       "dataset_size : 405\n",

  371.       "ave_node_num : 35.75061728395062\n",

  372.       "min_node_num : 13\n",

  373.       "max_node_num : 57\n",

  374.       "ave_edge_num : 38.358024691358025\n",

  375.       "min_edge_num : 13\n",

  376.       "max_edge_num : 60\n",

  377.       "ave_node_degree : 2.1466610247664697\n",

  378.       "min_node_degree : 2.0\n",

  379.       "max_node_degree : 2.2777777777777777\n",

  380.       "ave_fill_factor : 0.0314385616191916\n",

  381.       "min_fill_factor : 0.017851646660510926\n",

  382.       "max_fill_factor : 0.07692307692307693\n",

  383.       "node_label_num : 10\n",

  384.       "edge_label_num : 0\n",

  385.       "node_attr_dim : 3\n",

  386.       "edge_attr_dim : 0\n",

  387.       "class_number : 2\n",

  388.       "\n"

  389.      ]

  390.     },

  391.     {

  392.      "name": "stdout",

  393.      "output_type": "stream",

  394.      "text": [

  395.       "\n",

  396.       "COX2:\n",

  397.       "substructures : {'non linear', 'linear'}\n",

  398.       "node_labeled : True\n",

  399.       "edge_labeled : False\n",

  400.       "is_directed : False\n",

  401.       "dataset_size : 467\n",

  402.       "ave_node_num : 41.224839400428266\n",

  403.       "min_node_num : 32\n",

  404.       "max_node_num : 56\n",

  405.       "ave_edge_num : 43.44539614561028\n",

  406.       "min_edge_num : 34\n",

  407.       "max_edge_num : 59\n",

  408.       "ave_node_degree : 2.1077350079995685\n",

  409.       "min_node_degree : 2.076923076923077\n",

  410.       "max_node_degree : 2.1739130434782608\n",

  411.       "ave_fill_factor : 0.025799177869507202\n",

  412.       "min_fill_factor : 0.01881377551020408\n",

  413.       "max_fill_factor : 0.033203125\n",

  414.       "node_label_num : 8\n",

  415.       "edge_label_num : 0\n",

  416.       "node_attr_dim : 3\n",

  417.       "edge_attr_dim : 0\n",

  418.       "class_number : 2\n",

  419.       "\n",

  420.       "\n",

  421.       "DHFR:\n",

  422.       "substructures : {'non linear', 'linear'}\n",

  423.       "node_labeled : True\n",

  424.       "edge_labeled : False\n",

  425.       "is_directed : False\n",

  426.       "dataset_size : 756\n",

  427.       "ave_node_num : 42.42724867724868\n",

  428.       "min_node_num : 20\n",

  429.       "max_node_num : 71\n",

  430.       "ave_edge_num : 44.544973544973544\n",

  431.       "min_edge_num : 21\n",

  432.       "max_edge_num : 73\n",

  433.       "ave_node_degree : 2.102359895640024\n",

  434.       "min_node_degree : 2.0338983050847457\n",

  435.       "max_node_degree : 2.2\n",

  436.       "ave_fill_factor : 0.026126638866896944\n",

  437.       "min_fill_factor : 0.0144812537195001\n",

  438.       "max_fill_factor : 0.0525\n",

  439.       "node_label_num : 9\n",

  440.       "edge_label_num : 0\n",

  441.       "node_attr_dim : 3\n",

  442.       "edge_attr_dim : 0\n",

  443.       "class_number : 2\n",

  444.       "\n",

  445.       "\n",

  446.       "PROTEINS:\n",

  447.       "substructures : {'non linear', 'linear'}\n",

  448.       "node_labeled : True\n",

  449.       "edge_labeled : False\n",

  450.       "is_directed : False\n",

  451.       "dataset_size : 1113\n",

  452.       "ave_node_num : 39.05750224618149\n",

  453.       "min_node_num : 4\n",

  454.       "max_node_num : 620\n",

  455.       "ave_edge_num : 72.8158131176999\n",

  456.       "min_edge_num : 5\n",

  457.       "max_edge_num : 1049\n",

  458.       "ave_node_degree : 3.734642171150555\n",

  459.       "min_node_degree : 1.7142857142857142\n",

  460.       "max_node_degree : 5.071428571428571\n",

  461.       "ave_fill_factor : 0.09599853508460923\n",

  462.       "min_fill_factor : 0.0027289281997918834\n",

  463.       "max_fill_factor : 0.375\n",

  464.       "node_label_num : 3\n",

  465.       "edge_label_num : 0\n",

  466.       "node_attr_dim : 1\n",

  467.       "edge_attr_dim : 0\n",

  468.       "class_number : 2\n",

  469.       "\n",

  470.       "\n",

  471.       "PROTEINS_full:\n",

  472.       "substructures : {'non linear', 'linear'}\n",

  473.       "node_labeled : True\n",

  474.       "edge_labeled : False\n",

  475.       "is_directed : False\n",

  476.       "dataset_size : 1113\n",

  477.       "ave_node_num : 39.05750224618149\n",

  478.       "min_node_num : 4\n",

  479.       "max_node_num : 620\n",

  480.       "ave_edge_num : 72.8158131176999\n",

  481.       "min_edge_num : 5\n",

  482.       "max_edge_num : 1049\n",

  483.       "ave_node_degree : 3.734642171150555\n",

  484.       "min_node_degree : 1.7142857142857142\n",

  485.       "max_node_degree : 5.071428571428571\n",

  486.       "ave_fill_factor : 0.09599853508460923\n",

  487.       "min_fill_factor : 0.0027289281997918834\n",

  488.       "max_fill_factor : 0.375\n",

  489.       "node_label_num : 3\n",

  490.       "edge_label_num : 0\n",

  491.       "node_attr_dim : 29\n",

  492.       "edge_attr_dim : 0\n",

  493.       "class_number : 2\n",

  494.       "\n",

  495.       "\n",

  496.       "NCI1:\n",

  497.       "substructures : {'non linear', 'linear'}\n",

  498.       "node_labeled : True\n",

  499.       "edge_labeled : False\n",

  500.       "is_directed : False\n",

  501.       "dataset_size : 4110\n",

  502.       "ave_node_num : 29.8654501216545\n",

  503.       "min_node_num : 3\n",

  504.       "max_node_num : 111\n",

  505.       "ave_edge_num : 32.3\n",

  506.       "min_edge_num : 2\n",

  507.       "max_edge_num : 119\n",

  508.       "ave_node_degree : 2.155013792267071\n",

  509.       "min_node_degree : 0.8\n",

  510.       "max_node_degree : 2.769230769230769\n",

  511.       "ave_fill_factor : 0.04239828192835043\n",

  512.       "min_fill_factor : 0.009522961908152367\n",

  513.       "max_fill_factor : 0.2222222222222222\n",

  514.       "node_label_num : 37\n",

  515.       "edge_label_num : 0\n",

  516.       "node_attr_dim : 0\n",

  517.       "edge_attr_dim : 0\n",

  518.       "class_number : 2\n",

  519.       "\n",

  520.       "\n",

  521.       "NCI109:\n",

  522.       "substructures : {'non linear', 'linear'}\n",

  523.       "node_labeled : True\n",

  524.       "edge_labeled : False\n",

  525.       "is_directed : False\n",

  526.       "dataset_size : 4127\n",

  527.       "ave_node_num : 29.681124303368065\n",

  528.       "min_node_num : 4\n",

  529.       "max_node_num : 111\n",

  530.       "ave_edge_num : 32.13084565059365\n",

  531.       "min_edge_num : 3\n",

  532.       "max_edge_num : 119\n",

  533.       "ave_node_degree : 2.156446168619097\n",

  534.       "min_node_degree : 1.0909090909090908\n",

  535.       "max_node_degree : 2.769230769230769\n",

  536.       "ave_fill_factor : 0.04263668408405519\n",

  537.       "min_fill_factor : 0.009522961908152367\n",

  538.       "max_fill_factor : 0.1875\n",

  539.       "node_label_num : 38\n",

  540.       "edge_label_num : 0\n",

  541.       "node_attr_dim : 0\n",

  542.       "edge_attr_dim : 0\n",

  543.       "class_number : 2\n",

  544.       "\n",

  545.       "load SDF: 100%|██████████| 4457424/4457424 [00:08<00:00, 522501.84it/s]\n",

  546.       "ajust data: 100%|██████████| 42687/42687 [00:09<00:00, 4625.31it/s] \n",

  547.       "\n",

  548.       "NCI-HIV:\n",

  549.       "substructures : {'non linear', 'linear'}\n",

  550.       "node_labeled : True\n",

  551.       "edge_labeled : True\n",

  552.       "is_directed : False\n",

  553.       "dataset_size : 42682\n",

  554.       "ave_node_num : 45.70945597675835\n",

  555.       "min_node_num : 2\n",

  556.       "max_node_num : 438\n",

  557.       "ave_edge_num : 47.7137903565906\n",

  558.       "min_edge_num : 1\n",

  559.       "max_edge_num : 441\n",

  560.       "ave_node_degree : 2.087755727203458\n",

  561.       "min_node_degree : 1.0\n",

  562.       "max_node_degree : 4.571428571428571\n",

  563.       "ave_fill_factor : 0.02739985514266206\n",

  564.       "min_fill_factor : 0.002298742728466879\n",

  565.       "max_fill_factor : 0.25\n",

  566.       "node_label_num : 63\n",

  567.       "edge_label_num : 3\n",

  568.       "node_attr_dim : 0\n",

  569.       "edge_attr_dim : 0\n",

  570.       "class_number : 3\n",

  571.       "\n"

  572.      ]

  573.     }

  574.    ],

  575.    "source": [

  576.     "import sys\n",

  577.     "sys.path.insert(0, \"../\")\n",

  578.     "from pygraph.utils.graphfiles import loadDataset\n",

  579.     "from pygraph.utils.graphdataset import get_dataset_attributes\n",

  580.     "\n",

  581.     "dslist = [\n",

  582.     "    {'name': 'Acyclic', 'dataset': '../datasets/acyclic/dataset_bps.ds',},\n",

  583.     "    {'name': 'Alkane', 'dataset': '../datasets/Alkane/dataset.ds',\n",

  584.     "        'dataset_y': '../datasets/Alkane/dataset_boiling_point_names.txt',},\n",

  585.     "    {'name': 'MAO', 'dataset': '../datasets/MAO/dataset.ds',},\n",

  586.     "    {'name': 'PAH', 'dataset': '../datasets/PAH/dataset.ds',},\n",

  587.     "    {'name': 'MUTAG', 'dataset': '../datasets/MUTAG/MUTAG.mat',\n",

  588.     "        'extra_params': {'am_sp_al_nl_el': [0, 0, 3, 1, 2]}},\n",

  589.     "    {'name': 'Letter-med', 'dataset': '../datasets/Letter-med/Letter-med_A.txt'},\n",

  590.     "    {'name': 'ENZYMES', 'dataset': '../datasets/ENZYMES_txt/ENZYMES_A_sparse.txt'},\n",

  591.     "    {'name': 'Mutagenicity', 'dataset': '../datasets/Mutagenicity/Mutagenicity_A.txt'},\n",

  592.     "    {'name': 'D&D', 'dataset': '../datasets/D&D/DD.mat',\n",

  593.     "     'extra_params': {'am_sp_al_nl_el': [0, 1, 2, 1, -1]}},\n",

  594.     "    {'name': 'AIDS', 'dataset': '../datasets/AIDS/AIDS_A.txt'},\n",

  595.     "    {'name': 'FIRSTMM_DB', 'dataset': '../datasets/FIRSTMM_DB/FIRSTMM_DB_A.txt'},\n",

  596.     "    {'name': 'MSRC9', 'dataset': '../datasets/MSRC_9_txt/MSRC_9_A.txt'},\n",

  597.     "    {'name': 'MSRC21', 'dataset': '../datasets/MSRC_21_txt/MSRC_21_A.txt'},\n",

  598.     "    {'name': 'SYNTHETIC', 'dataset': '../datasets/SYNTHETIC_txt/SYNTHETIC_A_sparse.txt'},\n",

  599.     "    {'name': 'BZR', 'dataset': '../datasets/BZR_txt/BZR_A_sparse.txt'},\n",

  600.     "    {'name': 'COX2', 'dataset': '../datasets/COX2_txt/COX2_A_sparse.txt'},\n",

  601.     "    {'name': 'DHFR', 'dataset': '../datasets/DHFR_txt/DHFR_A_sparse.txt'},    \n",

  602.     "    {'name': 'PROTEINS', 'dataset': '../datasets/PROTEINS_txt/PROTEINS_A_sparse.txt'},\n",

  603.     "    {'name': 'PROTEINS_full', 'dataset': '../datasets/PROTEINS_full_txt/PROTEINS_full_A_sparse.txt'},   \n",

  604.     "    {'name': 'NCI1', 'dataset': '../datasets/NCI1/NCI1.mat',\n",

  605.     "        'extra_params': {'am_sp_al_nl_el': [1, 1, 2, 0, -1]}},\n",

  606.     "    {'name': 'NCI109', 'dataset': '../datasets/NCI109/NCI109.mat',\n",

  607.     "        'extra_params': {'am_sp_al_nl_el': [1, 1, 2, 0, -1]}},\n",

  608.     "    {'name': 'NCI-HIV', 'dataset': '../datasets/NCI-HIV/AIDO99SD.sdf',\n",

  609.     "        'dataset_y': '../datasets/NCI-HIV/aids_conc_may04.txt',},\n",

  610.     "\n",

  611.     "#     # not working below\n",

  612.     "#     {'name': 'PTC_FM', 'dataset': '../datasets/PTC/Train/FM.ds',},\n",

  613.     "#     {'name': 'PTC_FR', 'dataset': '../datasets/PTC/Train/FR.ds',},\n",

  614.     "#     {'name': 'PTC_MM', 'dataset': '../datasets/PTC/Train/MM.ds',},\n",

  615.     "#     {'name': 'PTC_MR', 'dataset': '../datasets/PTC/Train/MR.ds',},\n",

  616.     "]\n",

  617.     "\n",

  618.     "for ds in dslist:\n",

  619.     "    dataset, y = loadDataset(\n",

  620.     "        ds['dataset'],\n",

  621.     "        filename_y=(ds['dataset_y'] if 'dataset_y' in ds else None),\n",

  622.     "        extra_params=(ds['extra_params'] if 'extra_params' in ds else None))\n",

  623.     "    attrs = get_dataset_attributes(\n",

  624.     "        dataset, target=y, node_label='atom', edge_label='bond_type')\n",

  625.     "    print()\n",

  626.     "    print(ds['name'] + ':')\n",

  627.     "    for atr in attrs:\n",

  628.     "        print(atr, ':', attrs[atr])\n",

  629.     "    print()"

  630.    ]

  631.   }

  632.  ],

  633.  "metadata": {

  634.   "kernelspec": {

  635.    "display_name": "Python 3",

  636.    "language": "python",

  637.    "name": "python3"

  638.   },

  639.   "language_info": {

  640.    "codemirror_mode": {

  641.     "name": "ipython",

  642.     "version": 3

  643.    },

  644.    "file_extension": ".py",

  645.    "mimetype": "text/x-python",

  646.    "name": "python",

  647.    "nbconvert_exporter": "python",

  648.    "pygments_lexer": "ipython3",

  649.    "version": "3.6.7"

  650.   }

  651.  },

  652.  "nbformat": 4,

  653.  "nbformat_minor": 2

  654. }

A Python package for graph kernels, graph edit distances and graph pre-image problem.