Differences

This shows you the differences between two versions of the page.

--- papers:rg:ndungu2023 [2023/10/23 03:44] – asad
+++ papers:rg:ndungu2023 [2023/10/24 06:27] (current) – asad
@@ Line 7: / Line 7: @@
   - Need for classification: science, serendipitous discovery (Ray 2016).
   - Yatawatta 2021 gives a smart calibration package based on deep reinforcement learning.
+  - LOFAR sensitivity 100 $\mu$Jy, resolution $6''$.
+  - FR0 are 5 times more numerous than FR-I-II combined. The gradient in Fig.3 is nice.
+  - Is it right that WAT-NAT come from both FR I-II???
+  - IVOA (int virt obs alliance) gives FAIR (findable accessible interoperable reusable) data.
+  - Besides classification, need to focus on source extraction and anomaly detection. Create AI-alternatives to PyBDSF.
+  - Table 2 dos not have a dataset created from Sasmal 23.
+  - Most works depend on deep and shallow CNN (ConvNet, mimicks human visual cortex). The deeper (more layers) the better.
+  - Regularization techniques (dropout of weakly connected neurons) used to solve overfitting.
+  - Bowles 21 uses attention-gate layers to suppress irrelevant info.
+  - Tang 22 uses multidomain multibranch CNN to take multiple inputs.
+  - Brand 23 aligned the PC of all galaxies with the axis of the coordinate system via PCA.
+  - Sadeghi 21 uses Zernike polynomials (SVM) to extract image moments. Ntwaetsile 21 uses Haralick features for texture.
+  - Tang 19 successfully used FIRST-trained model on NVSS, but not the reverse.
+  - Wang 21 uses SKnet module for attention. Zhang 22 uses SE Net.
+  - Tiramisu (Pino 21) works best for source extraction, detection, localization and then classification.
+  - Main drawback of AI is the lack of explainability.
+  - Another drawback is that we are reducing 4D data cubes into 2D.