Positive-Congruent Training: Towards Regression-Free Model Updates

Reducing inconsistencies in the behavior of different versions of an AI system can be as important in practice as reducing its overall error. In image classification, sample-wise inconsistencies appear as "negative flips:" A new model incorrectly predicts the output for a test sample that was correctly classified by the old (reference) model. Positive-congruent (PC) training aims at reducing error rate while at the same time reducing negative flips, thus maximizing congruency with the reference model only on positive predictions, unlike model distillation. We propose a simple approach for PC training, Focal Distillation, which enforces congruence with the reference model by giving more weights to samples that were correctly classified. We also found that, if the reference model itself can be chosen as an ensemble of multiple deep neural networks, negative flips can be further reduced without affecting the new model's accuracy.

[1]  Stefan Wermter,et al.  Continual Lifelong Learning with Neural Networks: A Review , 2019, Neural Networks.

[2]  Natalia Gimelshein,et al.  PyTorch: An Imperative Style, High-Performance Deep Learning Library , 2019, NeurIPS.

[3]  Surya Ganguli,et al.  Continual Learning Through Synaptic Intelligence , 2017, ICML.

[4]  Robert E. Mercer,et al.  The Task Rehearsal Method of Life-Long Learning: Overcoming Impoverished Data , 2002, Canadian Conference on AI.

[5]  Xu Jia,et al.  Continual learning: A comparative study on how to defy forgetting in classification tasks , 2019, ArXiv.

[6]  Stefano Soatto,et al.  Towards Backward-Compatible Representation Learning , 2020, 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).

[7]  Chongruo Wu,et al.  ResNeSt: Split-Attention Networks , 2020, 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW).

[8]  Geoffrey E. Hinton,et al.  Distilling the Knowledge in a Neural Network , 2015, ArXiv.

[9]  Stefano Soatto,et al.  Entropy-SGD: biasing gradient descent into wide valleys , 2016, ICLR.

[10]  Bing Liu,et al.  Lifelong machine learning: a paradigm for continuous learning , 2017, Frontiers of Computer Science.

[11]  Philip H. S. Torr,et al.  Riemannian Walk for Incremental Learning: Understanding Forgetting and Intransigence , 2018, ECCV.

[12]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[13]  Yoshua Bengio,et al.  An Empirical Investigation of Catastrophic Forgeting in Gradient-Based Neural Networks , 2013, ICLR.

[14]  Derek Hoiem,et al.  Learning without Forgetting , 2016, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[15]  Marc'Aurelio Ranzato,et al.  Gradient Episodic Memory for Continual Learning , 2017, NIPS.

[16]  Yann LeCun,et al.  The Loss Surfaces of Multilayer Networks , 2014, AISTATS.

[17]  Yoshua Bengio,et al.  An Empirical Study of Example Forgetting during Deep Neural Network Learning , 2018, ICLR.

[18]  G. S. Watson.  Statistics on Spheres , 1983 .

[19]  Tinne Tuytelaars,et al.  A Continual Learning Survey: Defying Forgetting in Classification Tasks , 2019, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[20]  Stefano Soatto,et al.  Stochastic Gradient Descent Performs Variational Inference, Converges to Limit Cycles for Deep Networks , 2017, 2018 Information Theory and Applications Workshop (ITA).

[21]  R Ratcliff,et al.  Connectionist models of recognition memory: constraints imposed by learning and forgetting functions. , 1990, Psychological review.

[22]  Anderson Rocha,et al.  Toward Open Set Recognition , 2013, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[23]  Terrance E. Boult,et al.  Towards Open Set Deep Networks , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[24]  J. Kiefer,et al.  Stochastic Estimation of the Maximum of a Regression Function , 1952 .

[25]  Eric Horvitz,et al.  An Empirical Analysis of Backward Compatibility in Machine Learning Systems , 2020, KDD.

[26]  Yang Song,et al.  The iNaturalist Species Classification and Detection Dataset , 2017, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.

[27]  Sebastian Thrun,et al.  Lifelong Learning Algorithms , 1998, Learning to Learn.

[28]  Li Fei-Fei,et al.  ImageNet: A large-scale hierarchical image database , 2009, CVPR.

[29]  Kilian Q. Weinberger,et al.  Densely Connected Convolutional Networks , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[30]  Michael S. Bernstein,et al.  ImageNet Large Scale Visual Recognition Challenge , 2014, International Journal of Computer Vision.

[31]  Razvan Pascanu,et al.  Overcoming catastrophic forgetting in neural networks , 2016, Proceedings of the National Academy of Sciences.

[32]  Michael McCloskey,et al.  Catastrophic Interference in Connectionist Networks: The Sequential Learning Problem , 1989 .

[33]  Charles Blundell,et al.  Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles , 2016, NIPS.

[34]  Jian Sun,et al.  Deep Residual Learning for Image Recognition , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[35]  Philip H. S. Torr,et al.  GDumb: A Simple Approach that Questions Our Progress in Continual Learning , 2020, ECCV.

[36]  Christoph H. Lampert,et al.  iCaRL: Incremental Classifier and Representation Learning , 2016, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).