Rifki Kurniawan

Recent studies have demonstrated the potency of leveraging prompts in Transformers for continual learning (CL). Nevertheless, employing a discrete key-prompt bottleneck can lead to selection mismatches and inappropriate prompt associations during testing. Furthermore, this approach hinders adaptive prompting due to the lack of shareability among nearly identical instances at more granular level. To address these challenges, we introduce the Evolving Parameterized Prompt Memory (EvoPrompt), a novel method involving adaptive and continuous prompting attached to pre-trained Vision Transformer (ViT), conditioned on specific instance. We formulate a continuous prompt function as a neural bottleneck and encode the collection of prompts on network weights. We establish a paired prompt memory system consisting of a stable reference and a flexible working prompt memory. Inspired by linear mode connectivity, we progressively fuse the working prompt memory and reference prompt memory during inter-task periods, resulting in continually evolved prompt memory. This fusion involves aligning functionally equivalent prompts using optimal transport and aggregating them in parameter space with an adjustable bias based on prompt node attribution. Additionally, to enhance backward compatibility, we propose compositional classifier initialization, which leverages prior prototypes from pre-trained models to guide the initialization of new classifiers in a subspace-aware manner. Comprehensive experiments validate that our approach achieves state-of-the-art performance in both class and domain incremental learning scenarios.

Models pre-trained on the ImageNet dataset are introduced to be exploited for knowledge transfer in numerous downstream computer vision tasks, including the weakly-supervised anomaly detection and segmentation area. Specifically, in anomaly segmentation, the former study shows that representing images with feature maps extracted by pre-trained models significantly improves over previous techniques. This kind of representation method requires both high-quality and task-specific features, but feature extractors obtained from ImageNet directly are very general. One intuition for obtaining stronger features is by transferring a pre-trained model to the target dataset. However, in this paper, we show that under weakly-supervised settings, naïve fine-tune techniques that typically work for supervised learning can lead to catastrophic feature space collapse and reduce performance greatly. Thus, we propose to apply a topology-preserving constraint during transferring. Our method preserves the topology graph to keep the feature space from collapsing under weakly-supervised settings. And then we combine the transferred model with a simple anomaly detection and segmentation baseline for performance evaluation. The experiments show that our method achieves competitive accuracy on several benchmarks meanwhile setting a new state-of-the-art for anomaly detection on CIFAR100/10 and BTAD datasets.

Online continual learning in the wild is a very difficult task in machine learning. Non-stationarity in online continual learning potentially brings about catastrophic forgetting in neural networks. Specifically, online continual learning for autonomous driving with SODA10M dataset exhibits extra problems on extremely long-tailed distribution with continuous distribution shift. To address these problems, we propose multiple deep metric representation learning via both contrastive and supervised contrastive learning alongside soft labels distillation to improve model generalization. Moreover, we exploit modified class-balanced focal loss for sensitive penalization in class imbalanced and hard-easy samples. We also store some samples under guidance of uncertainty metric for rehearsal and perform online and periodical memory updates. Our proposed method achieves considerable generalization with average mean class accuracy (AMCA) 64.01% on validation and 64.53% AMCA on test set.