crowdkit.learning.conal.CoNAL | Source code
CoNAL(self,num_labels: int,n_workers: int,com_emb_size: int = 20,user_feature: Optional[...] = None)
Common Noise Adaptation Layers (CoNAL). This method introduces two types of confusions: worker-specific and
global. Each is parameterized by a confusion matrix. The ratio of the two confusions is determined by the common noise adaptation layer. The common noise adaptation layer is a trainable function that takes the instance embedding and the worker ID as input and outputs a scalar value between 0 and 1.
Zhendong Chu, Jing Ma, and Hongning Wang. Learning from Crowds by Modeling Common Confusions.
Proceedings of the AAAI Conference on Artificial Intelligence, 35(7), 5832-5840, 2021.
from crowdkit.learning import CoNALimport torchinput = torch.randn(3, 5)workers = torch.tensor([0, 1, 0])embeddings = torch.randn(3, 5)conal = CoNAL(5, 2)conal(embeddings, input, workers)
|forward||Forward pass of the CoNAL module.|
|simple_common_module||Common noise adoptation module.|
Last updated: March 31, 2023