Language-supervised vision models have recently attracted great attention in computer vision. A common approach to build such models is to use contrastive learning on paired data across the two modalities, as exemplified by Contrastive Language-Image Pre-Training (CLIP). In this paper, under linear representation settings, (i) we initiate the investigation of a general class of nonlinear loss functions for multimodal contrastive learning (MMCL) including CLIP loss and show its connection to singular value decomposition (SVD). Namely, we show that each step of loss minimization by gradient descent can be seen as performing SVD on a contrastive cross-covariance matrix. Based on this insight, (ii) we analyze the performance of MMCL. We quantitatively show that the feature learning ability of MMCL can be better than that of unimodal contrastive learning applied to each modality even under the presence of wrongly matched pairs. This characterizes the robustness of MMCL to noisy data. Furthermore, when we have access to additional unpaired data, (iii) we propose a new MMCL loss that incorporates additional unpaired datasets. We show that the algorithm can detect the ground-truth pairs and improve performance by fully exploiting unpaired datasets. The performance of the proposed algorithm was verified by numerical experiments.

We develop and analyze a principled approach to kernel ridge regression under covariate shift. The goal is to learn a regression function with small mean squared error over a target distribution, based on unlabeled data from there and labeled data that may have a different feature distribution. We propose to split the labeled data into two subsets and conduct kernel ridge regression on them separately to obtain a collection of candidate models and an imputation model. We use the latter to fill the missing labels and then select the best candidate model accordingly. Our non-asymptotic excess risk bounds show that in quite general scenarios, our estimator adapts to the structure of the target distribution as well as the covariate shift. It achieves the minimax optimal error rate up to a logarithmic factor. The use of pseudo-labels in model selection does not have major negative impacts.

End-to-end spoken language understanding (SLU) suffers from the long-tail word problem. This paper exploits contextual biasing, a technique to improve the speech recognition of rare words, in end-to-end SLU systems. Specifically, a tree-constrained pointer generator (TCPGen), a powerful and efficient biasing model component, is studied, which leverages a slot shortlist with corresponding entities to extract biasing lists. Meanwhile, to bias the SLU model output slot distribution, a slot probability biasing (SPB) mechanism is proposed to calculate a slot distribution from TCPGen. Experiments on the SLURP dataset showed consistent SLU-F1 improvements using TCPGen and SPB, especially on unseen entities. On a new split by holding out 5 slot types for the test, TCPGen with SPB achieved zero-shot learning with an SLU-F1 score over 50% compared to baselines which can not deal with it. In addition to slot filling, the intent classification accuracy was also improved.

Invariant Extended Kalman Filter (IEKF) has been successfully applied in Visual-inertial Odometry (VIO) as an advanced achievement of Kalman filter, showing great potential in sensor fusion. In this paper, we propose partial IEKF (PIEKF), which only incorporates rotation-velocity state into the Lie group structure and apply it for Visual-Inertial-Wheel Odometry (VIWO) to improve positioning accuracy and consistency. Specifically, we derive the rotation-velocity measurement model, which combines wheel measurements with kinematic constraints. The model circumvents the wheel odometer's 3D integration and covariance propagation, which is essential for filter consistency. And a plane constraint is also introduced to enhance the position accuracy. A dynamic outlier detection method is adopted, leveraging the velocity state output. Through the simulation and real-world test, we validate the effectiveness of our approach, which outperforms the standard Multi-State Constraint Kalman Filter (MSCKF) based VIWO in consistency and accuracy.

Data cubes are multidimensional databases, often built from several separate databases, that serve as flexible basis for data analysis. Surprisingly, outlier detection on data cubes has not yet been treated extensively. In this work, we provide the first framework to evaluate robust outlier detection methods in data cubes (RODD). We introduce a novel random forest-based outlier detection approach (RODD-RF) and compare it with more traditional methods based on robust location estimators. We propose a general type of test data and examine all methods in a simulation study. Moreover, we apply ROOD-RF to real world data. The results show that RODD-RF can lead to improved outlier detection.

Automated depression screening and diagnosis is a highly relevant problem today. There are a number of limitations of the traditional depression detection methods, namely, high dependence on clinicians and biased self-reporting. In recent years, research has suggested strong potential in machine learning (ML) based methods that make use of the user's passive data collected via wearable devices. However, ML is data hungry. Especially in the healthcare domain primary data collection is challenging. In this work, we present an approach based on transfer learning, from a model trained on a secondary dataset, for the real time deployment of the depression screening tool based on the actigraphy data of users. This approach enables machine learning modelling even with limited primary data samples. A modified version of leave one out cross validation approach performed on the primary set resulted in mean accuracy of 0.96, where in each iteration one subject's data from the primary set was set aside for testing.

Self-supervised learning (SSL) models confront challenges of abrupt informational collapse or slow dimensional collapse. We propose TriNet, which introduces a novel triple-branch architecture for preventing collapse and stabilizing the pre-training. TriNet learns the SSL latent embedding space and incorporates it to a higher level space for predicting pseudo target vectors generated by a frozen teacher. Our experimental results show that the proposed method notably stabilizes and accelerates pre-training and achieves a relative word error rate reduction (WERR) of 6.06% compared to the state-of-the-art (SOTA) Data2vec for a downstream benchmark ASR task. We will release our code at https://github.com/tencent-ailab/.

Out-of-distribution (OOD) inputs can compromise the performance and safety of real world machine learning systems. While many methods exist for OOD detection and work well on small scale datasets with lower resolution and few classes, few methods have been developed for large-scale OOD detection. Existing large-scale methods generally depend on maximum classification probability, such as the state-of-the-art grouped softmax method. In this work, we develop a novel approach that calculates the probability of the predicted class label based on label distributions learned during the training process. Our method performs better than current state-of-the-art methods with only a negligible increase in compute cost. We evaluate our method against contemporary methods across $14$ datasets and achieve a statistically significant improvement with respect to AUROC (84.2 vs 82.4) and AUPR (96.2 vs 93.7).

Detecting out-of-distribution (OOD) data is a task that is receiving an increasing amount of research attention in the domain of deep learning for computer vision. However, the performance of detection methods is generally evaluated on the task in isolation, rather than also considering potential downstream tasks in tandem. In this work, we examine selective classification in the presence of OOD data (SCOD). That is to say, the motivation for detecting OOD samples is to reject them so their impact on the quality of predictions is reduced. We show under this task specification, that existing post-hoc methods perform quite differently compared to when evaluated only on OOD detection. This is because it is no longer an issue to conflate in-distribution (ID) data with OOD data if the ID data is going to be misclassified. However, the conflation within ID data of correct and incorrect predictions becomes undesirable. We also propose a novel method for SCOD, Softmax Information Retaining Combination (SIRC), that augments softmax-based confidence scores with feature-agnostic information such that their ability to identify OOD samples is improved without sacrificing separation between correct and incorrect ID predictions. Experiments on a wide variety of ImageNet-scale datasets and convolutional neural network architectures show that SIRC is able to consistently match or outperform the baseline for SCOD, whilst existing OOD detection methods fail to do so.

In cities worldwide, cars cause health and traffic problems whichcould be partly mitigated through an increased modal share of bicycles. Many people, however, avoid cycling due to a lack of perceived safety. For city planners, addressing this is hard as they lack insights intowhere cyclists feel safe and where they do not. To gain such insights,we have in previous work proposed the crowdsourcing platform SimRa,which allows cyclists to record their rides and report near miss incidentsvia a smartphone app. In this paper, we present CycleSense, a combination of signal pro-cessing and Machine Learning techniques, which partially automatesthe detection of near miss incidents, thus making the reporting of nearmiss incidents easier. Using the SimRa data set, we evaluate CycleSenseby comparing it to a baseline method used by SimRa and show that itsignificantly improves incident detection.