Recent advances in Transformer models allow for unprecedented sequence lengths, due to linear space and time complexity. In the meantime, relative positional encoding (RPE) was proposed as beneficial for classical Transformers and consists in exploiting lags instead of absolute positions for inference. Still, RPE is not available for the recent linear-variants of the Transformer, because it requires the explicit computation of the attention matrix, which is precisely what is avoided by such methods. In this paper, we bridge this gap and present Stochastic Positional Encoding as a way to generate PE that can be used as a replacement to the classical additive (sinusoidal) PE and provably behaves like RPE. The main theoretical contribution is to make a connection between positional encoding and cross-covariance structures of correlated Gaussian processes. We illustrate the performance of our approach on the Long-Range Arena benchmark and on music generation.

---

Predicting extreme weather events such as tropical and extratropical cyclones is of vital scientific and societal importance. Of late, machine learning methods have found their way to weather analysis and prediction, but mostly, these methods use machine learning merely as a complement to traditional numerical weather prediction models. Although some pure machine learning and data-driven approaches for weather prediction have been developed, they mainly formulate the problem similar to pattern recognition or follow the train of thought of traditional time-series models for extreme weather event forecasting; for the former, this usually yields only single-step ahead prediction, and for the latter, this lacks the flexibility to account for observed weather features as such methods concern only the patterns of the extreme weather occurrences. In this paper, we depart from the typical practice of pattern recognition and time-series approaches and focus on employing machine learning to estimate the probabilities of extreme weather occurrences in a multi-step-ahead (MSA) fashion given information on both weather features and the realized occurrences of extreme weather. Specifically, we propose a Markov conditional forward (MCF) model that adopts the Markov property between the occurrences of extreme weather for MSA extreme weather forecasting. Moreover, for better long-term prediction, we propose three novel cube perturbation methods to address error accumulation in our model. Experimental results on a real-world extreme weather dataset show the superiority of the proposed MCF model in terms of prediction accuracy for both short-term and long-term forecasting; moreover, the three cube perturbation methods successfully increase the fault tolerance and generalization ability of the MCF model, yielding significant improvements for long-term prediction.

---

Device-free wireless indoor localization is a key enabling technology for the Internet of Things (IoT). Fingerprint-based indoor localization techniques are a commonly used solution. This paper proposes a semi-supervised, generative adversarial network (GAN)-based device-free fingerprinting indoor localization system. The proposed system uses a small amount of labeled data and a large amount of unlabeled data (i.e., semi-supervised), thus considerably reducing the expensive data labeling effort. Experimental results show that, as compared to the state-of-the-art supervised scheme, the proposed semi-supervised system achieves comparable performance with equal, sufficient amount of labeled data, and significantly superior performance with equal, highly limited amount of labeled data. Besides, the proposed semi-supervised system retains its performance over a broad range of the amount of labeled data. The interactions between the generator, discriminator, and classifier models of the proposed GAN-based system are visually examined and discussed. A mathematical description of the proposed system is also presented.

---

Current peripheral execution approaches for intermittently-powered systems require full access to the internal hardware state for checkpointing or rely on application-level energy estimation for task partitioning to make correct forward progress. Both requirements present significant practical challenges for energy-harvesting, intelligent edge IoT devices, which perform hardware accelerated DNN inference. Sophisticated compute peripherals may have inaccessible internal state, and the complexity of DNN models makes it difficult for programmers to partition the application into suitably sized tasks that fit within an estimated energy budget. This paper presents the concept of inference footprinting for intermittent DNN inference, where accelerator progress is accumulatively preserved across power cycles. Our middleware stack, HAWAII, tracks and restores inference footprints efficiently and transparently to make inference forward progress, without requiring access to the accelerator internal state and application-level energy estimation. Evaluations were carried out on a Texas Instruments device, under varied energy budgets and network workloads. Compared to a variety of task-based intermittent approaches, HAWAII improves the inference throughput by 5.7% to 95.7%, particularly achieving higher performance on heavily accelerated DNNs.

---

Textual data is common and informative auxiliary information for recommender systems. Most prior art utilizes text for rating predic- tion, but rare work connects it to top- N recommendation. Moreover, although advanced recommendation models capable of incorporating auxiliary information have been developed, none of these are specifically designed to model textual information, yielding a limited usage scenario for typical user-to-item recommendation. In this work, we present a framework of text-aware preference ranking (TPR) for top- N recommendation, in which we comprehensively model the joint association of user-item interaction and relations between items and associated text. Using the TPR framework, we construct a joint likelihood function that explicitly describes two ranking structures: 1) item preference ranking (IPR) and 2) word relatedness ranking (WRR), where the former captures the item preference of each user and the latter captures the word relatedness of each item. As these two explicit structures are by nature mutually dependent, we propose TPR-OPT, a simple yet effective learning criterion that additionally includes implicit structures, such as relatedness between items and relatedness between words for each user for model optimization. Such a design not only successfully describes the joint association among users, words, and text comprehensively but also naturally yields powerful representations that are suitable for a range of recommendation tasks, including user-to-item, item-to-item, and user-to-word recommendation, as well as item-to-word reconstruction. In this paper, extensive experiments have been conducted on eight recommendation datasets, the results of which demonstrate that by including textual information from item descriptions, the proposed TPR model consistently outperforms state-of-the-art baselines on various recommendation tasks.

---

Abstract—Sim-to-real, a term that describes where a model is trained in a simulator then transferred to the real world, is a technique that enables faster deep reinforcement learning (DRL) training. However, differences between the simulator and the real world often cause the model to perform poorly in the real world. Domain randomization is a way to bridge the sim-to-real gap by exposing the model to a wide range of scenarios so that it can generalize to real-world situations. However, following domain randomization to train an autonomous car racing model with DRL can lead to undesirable outcomes. Namely, a model trained with randomization tends to run slower; a higher completion rate on the testing track comes at the expense of longer lap times. This paper aims to boost the robustness of a trained race car model without compromising racing lap times. For a training track and a testing track having the same shape (and same optimal paths), but with different lighting, background, etc., we first train a model (teacher model) that overfits the training track, moving along a near optimal path. We then use this model to teach a student model the correct actions along with randomization. With our method, a model with 18.4% completion rate on the testing track is able to help teach a student model with 52% completion. Moreover, over an average of 50 trials, the student is able to finish a lap 0.23 seconds faster than the teacher. This 0.23 second gap is significant in tight races, with lap times of about 10 to 12 seconds. Index Terms—reinforcement learning, deep learning, miniature autonomous race car, sim-to-real, AWS DeepRacer

---

In recent years, the research community has approached the problem of vehicle re-identification (re-id) with attention-based models, specifically focusing on regions of a vehicle containing discriminative information. These re-id methods rely on expensive key-point labels, part annotations, and additional attributes including vehicle make, model, and color. Given the large number of vehicle re-id datasets with various levels of annotations, strongly-supervised methods are unable to scale across different domains. In this paper, we present Self-supervised Attention for Vehicle Re-identification (SAVER), a novel approach to effectively learn vehicle-specific discriminative features. Through extensive experimentation, we show that SAVER improves upon the state-of-the-art on challenging VeRi, VehicleID, Vehicle-1M and VERI-Wild datasets.

---

In this paper, we propose a novel optimization criterion that leverages features of the skew normal distribution to better model the problem of personalized recommendation. Specifically, the developed criterion borrows the concept and the flexibility of the skew normal distribution, based on which three hyperparameters are attached to the optimization criterion. Furthermore, from a theoretical point of view, we not only establish the relation between the maximization of the proposed criterion and the shape parameter in the skew normal distribution, but also provide the analogies and asymptotic analysis of the proposed criterion to maximization of the area under the ROC curve. Experimental results conducted on a range of large-scale real-world datasets show that our model significantly outperforms the state of the art and yields consistently best performance on all tested datasets.

---

Vector-valued neural learning has emerged as a promising direction in deep learning recently. Traditionally, training data for neural networks (NNs) are formulated as a vector of scalars; however, its performance may not be optimal since associations among adjacent scalars are not modeled. In this article, we propose a new vector neural architecture called the Arbitrary BIlinear Product NN (ABIPNN), which processes information as vectors in each neuron, and the feedforward projections are defined using arbitrary bilinear products. Such bilinear products can include circular convolution, 7-D vector product, skew circular convolution, reversed-time circular convolution, or other new products that are not seen in the previous work. As a proof-of-concept, we apply our proposed network to multispectral image denoising and singing voice separation. Experimental results show that ABIPNN obtains substantial improvements when compared to conventional NNs, suggesting that associations are learned during training.

---

Spatiotemporal super-resolution (SR) aims to upscale both the spatial and temporal dimensions of input videos, and produces videos with higher frame resolutions and rates. It involves two essential sub-tasks: spatial SR and temporal SR. We design a two-stream network for spatiotemporal SR in this work. One stream contains a temporal SR module followed by a spatial SR module, while the other stream has the same two modules in the reverse order. Based on the interchangeability of performing the two sub-tasks, the two network streams are supposed to produce consistent spatiotemporal SR results. Thus, we present a cross-stream consistency to enforce the similarity between the outputs of the two streams. In this way, the training of the two streams is correlated, which allows the two SR modules to share their supervisory signals and improve each other. In addition, the proposed cross-stream consistency does not consume labeled training data and can guide network training in an unsupervised manner. We leverage this property to carry out semi-supervised spatiotemporal SR. It turns out that our method makes the most of training data, and can derive an effective model with few high-resolution and high-frame-rate videos, achieving the state-of-the-art performance.

---