Anticipating future actions in a video is useful for many autonomous and assistive technologies. Most prior action anticipation work treat this as a vision modality problem, where the models learn the task information primarily from the video features in the action anticipation datasets. However, knowledge about action sequences can also be obtained from external textual data. In this work, we show how knowledge in pretrained language models can be adapted and distilled into vision-based action anticipation models. We Figure 1: A model learning the action anticipation from show that a simple distillation technique can only the vision modality (video frames) is essentially achieve effective knowledge transfer and provide exposed to a very limited set of action sequences. Language consistent gains on a strong vision model models, which are pre-trained on large-scale text, (Anticipative Vision Transformer) for two action can learn this distribution from the task, and a much anticipation datasets (3.5% relative gain larger domain-relevant text. We propose distilling this on EGTEA-GAZE+ and 7.2% relative gain on knowledge from text modality models to vision modality EPIC-KITCHEN 55), giving a new state-of-theart model for video action anticipation task.

A machine learning model that generalizes well should obtain low errors on unseen test examples. Thus, if we learn an optimal model in training data, it could have better generalization performance in testing tasks. However, learning such a model is not possible in standard machine learning frameworks as the distribution of the test data is unknown. To tackle this challenge, we propose a novel robust meta-learning method, which is more robust to the image-based testing tasks which is unknown and has distribution shifts with training tasks. Our robust meta-learning method can provide robust optimal models even when data from each distribution are scarce. In experiments, we demonstrate that our algorithm not only has better generalization performance but also robust to different unknown testing tasks.

In this work, we further develop the Physics-informed Spectral Learning (PiSL) by Espath et al. \cite{Esp21} based on a discrete $L^2$ projection to solve the discrete Hodge--Helmholtz decomposition from sparse data. Within this physics-informed statistical learning framework, we adaptively build a sparse set of Fourier basis functions with corresponding coefficients by solving a sequence of minimization problems where the set of basis functions is augmented greedily at each optimization problem. Moreover, our PiSL computational framework enjoys spectral (exponential) convergence. We regularize the minimization problems with the seminorm of the fractional Sobolev space in a Tikhonov fashion. In the Fourier setting, the divergence- and curl-free constraints become a finite set of linear algebraic equations. The proposed computational framework combines supervised and unsupervised learning techniques in that we use data concomitantly with the projection onto divergence- and curl-free spaces. We assess the capabilities of our method in various numerical examples including the `Storm of the Century' with satellite data from 1993.

Cross-sectional strategies are a classical and popular trading style, with recent high performing variants incorporating sophisticated neural architectures. While these strategies have been applied successfully to data-rich settings involving mature assets with long histories, deploying them on instruments with limited samples generally produce over-fitted models with degraded performance. In this paper, we introduce Fused Encoder Networks -- a novel and hybrid parameter-sharing transfer ranking model. The model fuses information extracted using an encoder-attention module operated on a source dataset with a similar but separate module focused on a smaller target dataset of interest. This mitigates the issue of models with poor generalisability that are a consequence of training on scarce target data. Additionally, the self-attention mechanism enables interactions among instruments to be accounted for, not just at the loss level during model training, but also at inference time. Focusing on momentum applied to the top ten cryptocurrencies by market capitalisation as a demonstrative use-case, the Fused Encoder Networks outperforms the reference benchmarks on most performance measures, delivering a three-fold boost in the Sharpe ratio over classical momentum as well as an improvement of approximately 50% against the best benchmark model without transaction costs. It continues outperforming baselines even after accounting for the high transaction costs associated with trading cryptocurrencies.

In recent years, there has been a growing interest in the development of language models capable of generating text with controllable attributes. While several approaches have been proposed, many of these methods require condition-specific data or significant computational resources. In this study, we propose a novel method called Gamma Sampling, which enables controllable language generation without the need for any training data and maintains a fast generation speed. Gamma Sampling incorporates attribute-related information into the sampling process, effectively guiding the language model to produce text with desired attributes. Our experimental results demonstrate that Gamma Sampling, when applied to GPT2, outperforms representative baselines in terms of diversity, attribute relevance, and overall quality of the generated samples.

Density Ratio Estimation (DRE) is an important machine learning technique with many downstream applications. We consider the challenge of DRE with missing not at random (MNAR) data. In this setting, we show that using standard DRE methods leads to biased results while our proposal (M-KLIEP), an adaptation of the popular DRE procedure KLIEP, restores consistency. Moreover, we provide finite sample estimation error bounds for M-KLIEP, which demonstrate minimax optimality with respect to both sample size and worst-case missingness. We then adapt an important downstream application of DRE, Neyman-Pearson (NP) classification, to this MNAR setting. Our procedure both controls Type I error and achieves high power, with high probability. Finally, we demonstrate promising empirical performance both synthetic data and real-world data with simulated missingness.

Explainable artificial intelligence is increasingly used in machine learning (ML) based decision-making systems in healthcare. However, little research has compared the utility of different explanation methods in guiding healthcare experts for patient care. Moreover, it is unclear how useful, understandable, actionable and trustworthy these methods are for healthcare experts, as they often require technical ML knowledge. This paper presents an explanation dashboard that predicts the risk of diabetes onset and explains those predictions with data-centric, feature-importance, and example-based explanations. We designed an interactive dashboard to assist healthcare experts, such as nurses and physicians, in monitoring the risk of diabetes onset and recommending measures to minimize risk. We conducted a qualitative study with 11 healthcare experts and a mixed-methods study with 45 healthcare experts and 51 diabetic patients to compare the different explanation methods in our dashboard in terms of understandability, usefulness, actionability, and trust. Results indicate that our participants preferred our representation of data-centric explanations that provide local explanations with a global overview over other methods. Therefore, this paper highlights the importance of visually directive data-centric explanation method for assisting healthcare experts to gain actionable insights from patient health records. Furthermore, we share our design implications for tailoring the visual representation of different explanation methods for healthcare experts.

Transformers have demonstrated impressive strength in long-term series forecasting. Existing prediction research mostly focused on mapping past short sub-series (lookback window) to future series (forecast window). The longer training dataset time series will be discarded, once training is completed. Models can merely rely on lookback window information for inference, which impedes models from analyzing time series from a global perspective. And these windows used by Transformers are quite narrow because they must model each time-step therein. Under this point-wise processing style, broadening windows will rapidly exhaust their model capacity. This, for fine-grained time series, leads to a bottleneck in information input and prediction output, which is mortal to long-term series forecasting. To overcome the barrier, we propose a brand-new methodology to utilize Transformer for time series forecasting. Specifically, we split time series into patches by day and reform point-wise to patch-wise processing, which considerably enhances the information input and output of Transformers. To further help models leverage the whole training set's global information during inference, we distill the information, store it in time representations, and replace series with time representations as the main modeling entities. Our designed time-modeling Transformer -- Dateformer yields state-of-the-art accuracy on 7 real-world datasets with a 33.6\% relative improvement and extends the maximum forecast range to half-year.

Language contact is a pervasive phenomenon reflected in the borrowing of words from donor to recipient languages. Most computational approaches to borrowing detection treat all languages under study as equally important, even though dominant languages have a stronger impact on heritage languages than vice versa. We test new methods for lexical borrowing detection in contact situations where dominant languages play an important role, applying two classical sequence comparison methods and one machine learning method to a sample of seven Latin American languages which have all borrowed extensively from Spanish. All methods perform well, with the supervised machine learning system outperforming the classical systems. A review of detection errors shows that borrowing detection could be substantially improved by taking into account donor words with divergent meanings from recipient words.

In the last two decades, the combination of machine learning and quantum computing has been an ever-growing topic of interest but, to this date, the limitations of quantum computing hardware have somewhat restricted the use of complex multi-qubit operations for machine learning. In this paper, we capitalize on the cyclical nature of quantum state probabilities observed on pulsed qubits to propose a single-qubit feed forward block whose architecture allows for classical parameters to be used in a way similar to classical neural networks. To do this, we modulate the pulses exciting qubits to induce superimposed rotations around the Bloch Sphere. The approach presented here has the advantage of employing a single qubit per block. Thus, it is linear with respect to the number of blocks, not polynomial with respect to the number of neurons as opposed to the majority of methods elsewhere. Further, since it employs classical parameters, a large number of iterations and updates at training can be effected without dwelling on coherence times and the gradients can be reused and stored if necessary. We also show how an analogy can be drawn to neural networks using sine-squared activation functions and illustrate how the feed-forward block presented here may be used and implemented on pulse-enabled quantum computers.