【AAAI 2021】Few-Shot One-Class Classification via Meta-Learning 【FSOCC via Meta-learning】

Background knowledge

One-class classification

learning a binary classifier with data from only one class // Learn a two classifier from a class . 【 idea 】 Can be used for anomaly detection .

The anomaly detection (AD) task

(Chandola, Banerjee, and Kumar 2009; Aggarwal 2015) consists in differentiating between normal and abnormal data samples. AD
AD problems are usually formulated as one-class classification (OCC) problems (Moya, Koch, and Hostetler 1993), where either only a few or no anomalous data samples are available for training the model (Khan and Madden 2014).

Contents summary

Our work addresses the few-shot OCC problem and presents a method to modify the episodic data sampling strategy of the model-agnostic meta-learning (MAML) algorithm to learn a model initialization particularly suited for learning few-shot OCC tasks. // Our work has solved few-shot OCC problem , A method to modify model independent meta learning is proposed (MAML) Scenario data sampling strategy of the algorithm , Learning one is especially suitable for learning few-shot OCC Task model initialization .

We provide a theoretical analysis that explains why our approach works in the few-shot OCC scenario, while other meta-learning al- gorithms fail, including the unmodified MAML. // We provide a theoretical analysis , Explains why our approach is few-shot OCC Effective in the scene , Other meta learning algorithms fail , Include unmodified MAML.

Our exper- iments on eight datasets from the image and time-series do- mains show that our method leads to better results than classi- cal OCC and few-shot classification approaches, and demon- strate the ability to learn unseen tasks from only few nor- mal class samples. // We are 8 Experiments on image and time series data sets show that , Our method is better than classical OCC and few-shot The classification method achieves better results , And proved the ability to learn invisible tasks from only a few ordinary samples

we successfully train anomaly detectors for a real-world application on sensor readings recorded during industrial manufacturing of workpieces with a CNC milling machine, by using few normal examples. // By using a few common examples , We have successfully trained the anomaly detector , In order to apply it to CNC Sensor readings recorded during industrial manufacturing of milling machine workpiece .

contributions

1. Firstly, we show that classical OCC approaches fail in the few-shot data regime. We proved the classic OCC Method in few-shot Failed in data .
2. Secondly, we provide a theoretical analysis showing that classical gradient-based meta-learning algorithms do not yield parameter initializations suitable for OCC and that second- order derivatives are needed to optimize for such initializations. We Provides a theoretical analysis , It shows that the classical gradient based meta learning algorithm No fit OCC Parameter initialization of , also Second derivative This initialization needs to be optimized .
3. Thirdly, we propose a simple episode generation strat- egy to adapt any meta-learning algorithm that uses a bi-level optimization
   scheme to FS-OCC. Hereby, we first focus on modifying the model-agnostic meta-learning (MAML) al- gorithm (Finn, Abbeel, and Levine 2017) to learn initializa- tions useful for the FS-OCC scenario. The resulting One- Class MAML (OC-MAML) maximizes the inner product of loss gradients computed on one-class and class-balanced minibatches, hence maximizing the cosine similarity be- tween these gradients.
4. Finally, we demonstrate that the pro- posed data sampling technique generalizes beyond MAML to other metalearning algorithms, e.g., MetaOptNet (Lee et al. 2019) and Meta-SGD (Li et al. 2017), by successfully adapting them to the understudied FS-OCC.

The main method

Algorithm description