Hi, is it possible that the number of dataset is 100 and not 30? What is the meaning of target_num, label_num fields? Thanks.
Posted by: AIpert @ Jan. 18, 2022, 6:31 p.m.Hi AIpert,
The sample data included in the starting kit includes synthetic learning curves of 20 algorithms on 100 datasets, which is completely different from the data used for testing and ranking participants in our competition.
Kindly find the dataset meta-features and algorithm meta-features information at the end of this message. More details can be found in the "Get Data" tab (click on "Participate" -> "Get Data").
Best regards,
The organizing team
===================
DATASET META-FEATURES
* usage = 'AutoML challenge 2014'
* name = name of the dataset
* task = ’binary.classification’, ’multiclass.classification’, ’multilabel.classification’, ’regression’
* target_type = ’Binary’, ’Categorical’, ’Numerical’
* feat_type = ’Binary’, ’Categorical’, ’Numerical’
* metric = ’bac’, ’auc’, ’f1’, ’pac’, ’abs’, ’r2’
* time_budget = total time budget for running algorithms on the dataset
* feat_num = number of features
* target_num = number of columns of target file (one, except for multi-label problems)
* label_num = number of labels (number of unique values of the targets)
* train_num = number of training examples
* valid_num = number of validation examples
* test_num = number of test examples
* has_categorical = whether there are categorical variable (yes=1, no=0)
* has_missing = whether there are missing values (yes=1, no=0)
* is_sparse = whether this is a sparse dataset (yes=1, no=0)
ALGORITHM META-FEATURES
* meta_feature_0 = 0 or 1
* meta_feature_1 = 0.1, 0.2, 0.3,…, 1.0
Thank you for your answer.
I should point out that 'feat_type'='Mixed' that is present in the provided sample data is not mentioned anywhere.
Maybe there is a typo in the provided description of the dataset's meta-features?
Posted by: AIpert @ Jan. 24, 2022, 5:38 p.m.Hi,
We apologize for the typos. The correct value set of feat_type is:
feat_type = 'Binary', 'Categorical', 'Numerical', 'Mixed'
Thank you very much for pointing it out.
Best regards,
Hung
Hi,
We fixed the typos and updated the starting kit.
Please download the latest starting kit and use it for developing your agent.
Best regards,
Hung