【PAPER-summary】Papers read 2023
papers that I’ve read in 2023.
last edited 2023.8.11
August
8.9 Causality for Machine Learning
1
some of causality works is already in the process of entering the machine learning mainstream, in particular the view that causal modeling can lead to more invariant or robust models.
Some think we are now in the middle of third revolution of information, which is also likely a conserved quantity that could only be converted or harvested rather than generating it from air.The first industrial revolutions rendered energy a universal currency; the same may be happening to information.
One can argue that the present revolution has two components:
- Computer Science and AI manipulatiing symbols
- About learning,extract information also from unstructured data
In recent years, genuine connections between machine learning and causality have emerged, and we will argue that these connections are crucial if we want to make progress on the major open problems of AI.
2
2023.8.11 Lung nodule detection algorithm based on rank correlation causal structure learning
combine causal inference and machine learning in order to classify lung nodules.
The whole process is defined in three parts:
- First they obtain possible neighbor nodes(skeletons of the causal network) for both computed feature and semantic feature nodes from the input lung nodule dataset.
- Use score based methods and algorithms to get the final causal structure network.
- Feature selection based on the causal structure network diagram, build a subset of features by selecting computational feature nodes related to semantic fearure prediction.And train a calssifier with this new dataset to predict the level of semantic features.
2023.8.11 Toward Causal Representation Learning
From Fig.2. and Fig.3.,we can tell a relation that even a sparse change of causal variable can lead to dense like pixel level change.