AETTA: Label-Free Accuracy Estimation for Test-Time Adaptation

Test-time adaptation (TTA) has emerged as a viable solution to adapt pre-trained models to domain shifts using unlabeled test data. However, TTA faces challenges of adaptation failures due to its reliance on blind adaptation to unknown test samples in dynamic scenarios. Traditional methods for out-of-distribution performance estimation are limited by unrealistic assumptions in the TTA context, such as requiring labeled data or re-training models. To address this issue, we propose AETTA, a label-free accuracy estimation algorithm for TTA. We propose the prediction disagreement as the accuracy estimate, calculated by comparing the target model prediction with dropout inferences. We then improve the prediction disagreement to extend the applicability of AETTA under adaptation failures. Our extensive evaluation with four baselines and six TTA methods demonstrates that AETTA shows an average of 19.8%p more accurate estimation compared with the baselines. We further demonstrate the effectiveness of accuracy estimation with a model recovery case study, showcasing the practicality of our model recovery based on accuracy estimation. The source code is available at https://github.com/taeckyung/AETTA.

SoTTA: Robust Test-Time Adaptation on Noisy Data Streams

Test-time adaptation (TTA) aims to address distributional shifts between training and testing data using only unlabeled test data streams for continual model adaptation. However, most TTA methods assume benign test streams, while test samples could be unexpectedly diverse in the wild. For instance, an unseen object or noise could appear in autonomous driving. This leads to a new threat to existing TTA algorithms; we found that prior TTA algorithms suffer from those noisy test samples as they blindly adapt to incoming samples. To address this problem, we present Screening-out Test-Time Adaptation (SoTTA), a novel TTA algorithm that is robust to noisy samples. The key enabler of SoTTA is two-fold: (i) input-wise robustness via high-confidence uniform-class sampling that effectively filters out the impact of noisy samples and (ii) parameter-wise robustness via entropy-sharpness minimization that improves the robustness of model parameters against large gradients from noisy samples. Our evaluation with standard TTA benchmarks with various noisy scenarios shows that our method outperforms state-of-the-art TTA methods under the presence of noisy samples and achieves comparable accuracy to those methods without noisy samples. The source code is available at https://github.com/taeckyung/SoTTA .

LanSER: Language-Model Supported Speech Emotion Recognition

Speech emotion recognition (SER) models typically rely on costly human-labeled data for training, making scaling methods to large speech datasets and nuanced emotion taxonomies difficult. We present LanSER, a method that enables the use of unlabeled data by inferring weak emotion labels via pre-trained large language models through weakly-supervised learning. For inferring weak labels constrained to a taxonomy, we use a textual entailment approach that selects an emotion label with the highest entailment score for a speech transcript extracted via automatic speech recognition. Our experimental results show that models pre-trained on large datasets with this weak supervision outperform other baseline models on standard SER datasets when fine-tuned, and show improved label efficiency. Despite being pre-trained on labels derived only from text, we show that the resulting representations appear to model the prosodic content of speech.

IMG2IMU: Applying Knowledge from Large-Scale Images to IMU Applications via Contrastive Learning

Recent advances in machine learning showed that pre-training representations acquired via self-supervised learning could achieve high accuracy on tasks with small training data. Unlike in vision and natural language processing domains, such pre-training for IMU-based applications is challenging, as there are only a few publicly available datasets with sufficient size and diversity to learn generalizable representations. To overcome this problem, we propose IMG2IMU, a novel approach that adapts pre-train representation from large-scale images to diverse few-shot IMU sensing tasks. We convert the sensor data into visually interpretable spectrograms for the model to utilize the knowledge gained from vision. Further, we apply contrastive learning on an augmentation set we designed to learn representations that are tailored to interpreting sensor data. Our extensive evaluations on five different IMU sensing tasks show that IMG2IMU consistently outperforms the baselines, illustrating that vision knowledge can be incorporated into a few-shot learning environment for IMU sensing tasks.

Robust Continual Test-time Adaptation: Instance-aware BN and Prediction-balanced Memory

Test-time adaptation (TTA) is an emerging paradigm that addresses distributional shifts between training and testing phases without additional data acquisition or labeling cost; only unlabeled test data streams are used for continual model adaptation. Previous TTA schemes assume that the test samples are independent and identically distributed (i.i.d.), even though they are often temporally correlated (non-i.i.d.) in application scenarios, e.g., autonomous driving. We discover that most existing TTA methods fail dramatically under such scenarios. Motivated by this, we present a new test-time adaptation scheme that is robust against non-i.i.d. test data streams. Our novelty is mainly two-fold: (a) Instance-Aware Batch Normalization (IABN) that corrects normalization for out-of-distribution samples, and (b) Prediction-balanced Reservoir Sampling (PBRS) that simulates i.i.d. data stream from non-i.i.d. stream in a class-balanced manner. Our evaluation with various datasets, including real-world non-i.i.d. streams, demonstrates that the proposed robust TTA not only outperforms state-of-the-art TTA algorithms in the non-i.i.d. setting, but also achieves comparable performance to those algorithms under the i.i.d. assumption.

DAPPER: Performance Estimation of Domain Adaptation in Mobile Sensing

Many applications that utilize sensors in mobile devices and apply machine learning to provide novel services have emerged. However, various factors such as different users, devices, environments, and hyperparameters, affect the performance for such applications, thus making the domain shift (i.e., distribution shift of a target user from the training source dataset) an important problem. Although recent domain adaptation techniques attempt to solve this problem, the complex interplay between the diverse factors often limits their effectiveness. We argue that accurately estimating the performance in untrained domains could significantly reduce performance uncertainty. We present DAPPER (Domain AdaPtation Performance EstimatoR) that estimates the adaptation performance in a target domain with only unlabeled target data. Our intuition is that the outputs of a model on the target data provide clues for the model's actual performance in the target domain. DAPPER does not require expensive labeling costs nor involve additional training after deployment. Our evaluation with four real-world sensing datasets compared against four baselines shows that DAPPER outperforms the baselines by on average 17% in estimation accuracy. Moreover, our on-device experiment shows that DAPPER achieves up to 216X less computation overhead compared with the baselines.