Free Access
Table 9.
Testing accuracy of the selected algorithms for the three-class classification of 3FGL sources.
Algorithm | Parameters | Testing | Std. Dev. | Comparison with |
---|---|---|---|---|
accuracy | 4FGL-DR2 accuracy | |||
RF | 50 trees, max depth 6 | 93.96 | 0.85 | 85.00 |
RF_O | 94.38 | 0.76 | 85.00 | |
BDT | 100 trees, max depth 2 | 93.72 | 0.83 | 83.24 |
BDT_O | 93.83 | 0.80 | 85.29 | |
NN | 600 epochs, 11 neurons, LBFGS | 93.17 | 1.05 | 83.53 |
NN_O | 92.51 | 1.34 | 81.76 | |
LR | 500 iterations, LBFGS solver | 93.93 | 0.88 | 83.24 |
LR_O | 93.01 | 0.96 | 83.24 |
Notes. Comparison with associations in the 4FGL-DR2 catalog is presented in the last column. “_O” denotes training with oversampling.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.