Timezone: »

To Smooth or Not? When Label Smoothing Meets Noisy Labels
Jiaheng Wei · Hangyu Liu · Tongliang Liu · Gang Niu · Masashi Sugiyama · Yang Liu

Tue Jul 19 03:30 PM -- 05:30 PM (PDT) @ Hall E #215

Label smoothing (LS) is an arising learning paradigm that uses the positively weighted average of both the hard training labels and uniformly distributed soft labels. It was shown that LS serves as a regularizer for training data with hard labels and therefore improves the generalization of the model. Later it was reported LS even helps with improving robustness when learning with noisy labels. However, we observed that the advantage of LS vanishes when we operate in a high label noise regime. Intuitively speaking, this is due to the increased entropy of P(noisy label|X) when the noise rate is high, in which case, further applying LS tends to “over-smooth” the estimated posterior. We proceeded to discover that several learning-with-noisy-labels solutions in the literature instead relate more closely to not/negative label smoothing (NLS), which acts counter to LS and defines as using a negative weight to combine the hard and soft labels! We provide understandings for the properties of LS and NLS when learning with noisy labels. Among other established properties, we theoretically show NLS is considered more beneficial when the label noise rates are high. We provide extensive experimental results on multiple benchmarks to support our findings too.

Author Information

Jiaheng Wei (University of California, Santa Cruz)
Hangyu Liu (Brown University)
Tongliang Liu (The University of Sydney)
Gang Niu (RIKEN)
Gang Niu

Gang Niu is currently an indefinite-term research scientist at RIKEN Center for Advanced Intelligence Project.

Masashi Sugiyama (RIKEN / The University of Tokyo)
Yang Liu (UC Santa Cruz)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors