Calibration of Continual Learning Models
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2024
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866916203160666112 |
|---|---|
| author | Li, Lanpei Piccoli, Elia Cossu, Andrea Bacciu, Davide Lomonaco, Vincenzo |
| author_facet | Li, Lanpei Piccoli, Elia Cossu, Andrea Bacciu, Davide Lomonaco, Vincenzo |
| contents | Continual Learning (CL) focuses on maximizing the predictive performance of a model across a non-stationary stream of data. Unfortunately, CL models tend to forget previous knowledge, thus often underperforming when compared with an offline model trained jointly on the entire data stream. Given that any CL model will eventually make mistakes, it is of crucial importance to build calibrated CL models: models that can reliably tell their confidence when making a prediction. Model calibration is an active research topic in machine learning, yet to be properly investigated in CL. We provide the first empirical study of the behavior of calibration approaches in CL, showing that CL strategies do not inherently learn calibrated models. To mitigate this issue, we design a continual calibration approach that improves the performance of post-processing calibration methods over a wide range of different benchmarks and CL strategies. CL does not necessarily need perfect predictive models, but rather it can benefit from reliable predictive models. We believe our study on continual calibration represents a first step towards this direction. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2404_07817 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Calibration of Continual Learning Models Li, Lanpei Piccoli, Elia Cossu, Andrea Bacciu, Davide Lomonaco, Vincenzo Machine Learning Artificial Intelligence Continual Learning (CL) focuses on maximizing the predictive performance of a model across a non-stationary stream of data. Unfortunately, CL models tend to forget previous knowledge, thus often underperforming when compared with an offline model trained jointly on the entire data stream. Given that any CL model will eventually make mistakes, it is of crucial importance to build calibrated CL models: models that can reliably tell their confidence when making a prediction. Model calibration is an active research topic in machine learning, yet to be properly investigated in CL. We provide the first empirical study of the behavior of calibration approaches in CL, showing that CL strategies do not inherently learn calibrated models. To mitigate this issue, we design a continual calibration approach that improves the performance of post-processing calibration methods over a wide range of different benchmarks and CL strategies. CL does not necessarily need perfect predictive models, but rather it can benefit from reliable predictive models. We believe our study on continual calibration represents a first step towards this direction. |
| title | Calibration of Continual Learning Models |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2404.07817 |