Motivated by the developing mathematics of deep learning, we build universal functions approximators of continuous maps between arbitrary Polish metric spaces $\mathcal{X}$ and $\mathcal{Y}$ using elementary functions between Euclidean spaces as building blocks. Earlier results assume that the target space $\mathcal{Y}$ is a topological vector space. We overcome this limitation by ``randomization'': our approximators output discrete probability measures over $\mathcal{Y}$. When $\mathcal{X}$ and $\mathcal{Y}$ are Polish without additional structure, we prove very general qualitative guarantees; when they have suitable combinatorial structure, we prove quantitative guarantees for H\"{o}lder-like maps, including maps between finite graphs, solution operators to rough differential equations between certain Carnot groups, and continuous non-linear operators between Banach spaces arising in inverse problems. In particular, we show that the required number of Dirac measures is determined by the combinatorial structure of $\mathcal{X}$ and $\mathcal{Y}$. For barycentric $\mathcal{Y}$, including Banach spaces, $\mathbb{R}$-trees, Hadamard manifolds, or Wasserstein spaces on Polish metric spaces, our approximators reduce to $\mathcal{Y}$-valued functions. When the Euclidean approximators are neural networks, our constructions generalize transformer networks, providing a new probabilistic viewpoint of geometric deep learning.

Models based on bidirectional encoder representations from transformers (BERT) produce state of the art (SOTA) results on many natural language processing (NLP) tasks such as named entity recognition (NER), part-of-speech (POS) tagging etc. An interesting phenomenon occurs when classifying long documents such as those from the US supreme court where BERT-based models can be considered difficult to use on a first-pass or out-of-the-box basis. In this paper, we experiment with several BERT-based classification techniques for US supreme court decisions or supreme court database (SCDB) and compare them with the previous SOTA results. We then compare our results specifically with SOTA models for long documents. We compare our results for two classification tasks: (1) a broad classification task with 15 categories and (2) a fine-grained classification task with 279 categories. Our best result produces an accuracy of 80\% on the 15 broad categories and 60\% on the fine-grained 279 categories which marks an improvement of 8\% and 28\% respectively from previously reported SOTA results.

Domain gaps are among the most relevant roadblocks in the clinical translation of machine learning (ML)-based solutions for medical image analysis. While current research focuses on new training paradigms and network architectures, little attention is given to the specific effect of prevalence shifts on an algorithm deployed in practice. Such discrepancies between class frequencies in the data used for a method's development/validation and that in its deployment environment(s) are of great importance, for example in the context of artificial intelligence (AI) democratization, as disease prevalences may vary widely across time and location. Our contribution is twofold. First, we empirically demonstrate the potentially severe consequences of missing prevalence handling by analyzing (i) the extent of miscalibration, (ii) the deviation of the decision threshold from the optimum, and (iii) the ability of validation metrics to reflect neural network performance on the deployment population as a function of the discrepancy between development and deployment prevalence. Second, we propose a workflow for prevalence-aware image classification that uses estimated deployment prevalences to adjust a trained classifier to a new environment, without requiring additional annotated deployment data. Comprehensive experiments based on a diverse set of 30 medical classification tasks showcase the benefit of the proposed workflow in generating better classifier decisions and more reliable performance estimates compared to current practice.

To enhance the quality of generated stories, recent story generation models have been investigating the utilization of higher-level attributes like plots or commonsense knowledge. The application of prompt-based learning with large language models (LLMs), exemplified by GPT-3, has exhibited remarkable performance in diverse natural language processing (NLP) tasks. This paper conducts a comprehensive investigation, utilizing both automatic and human evaluation, to compare the story generation capacity of LLMs with recent models across three datasets with variations in style, register, and length of stories. The results demonstrate that LLMs generate stories of significantly higher quality compared to other story generation models. Moreover, they exhibit a level of performance that competes with human authors, albeit with the preliminary observation that they tend to replicate real stories in situations involving world knowledge, resembling a form of plagiarism.

Advances in deep learning for human activity recognition have been relatively limited due to the lack of large labelled datasets. In this study, we leverage self-supervised learning techniques on the UK-Biobank activity tracker dataset--the largest of its kind to date--containing more than 700,000 person-days of unlabelled wearable sensor data. Our resulting activity recognition model consistently outperformed strong baselines across seven benchmark datasets, with an F1 relative improvement of 2.5%-100% (median 18.4%), the largest improvements occurring in the smaller datasets. In contrast to previous studies, our results generalise across external datasets, devices, and environments. Our open-source model will help researchers and developers to build customisable and generalisable activity classifiers with high performance.

We develop concentration inequalities for the $l_\infty$ norm of a vector linear processes on mixingale sequences with sub-Weibull tails. These inequalities make use of the Beveridge-Nelson decomposition, which reduces the problem to concentration for sup-norm of a vector-mixingale or its weighted sum. This inequality is used to obtain a concentration bound for the maximum entrywise norm of the lag-$h$ autocovariance matrices of linear processes. These results are useful for estimation bounds for high-dimensional vector-autoregressive processes estimated using $l_1$ regularisation, high-dimensional Gaussian bootstrap for time series, and long-run covariance matrix estimation.

Numerous examples in the literature proved that deep learning models have the ability to work well with multimodal data. Recently, CLIP has enabled deep learning systems to learn shared latent spaces between images and text descriptions, with outstanding zero- or few-shot results in downstream tasks. In this paper we explore the same idea proposed by CLIP but applied to the speech domain, where the phonetic and acoustic spaces usually coexist. We train a CLIP-based model with the aim to learn shared representations of phonetic and acoustic spaces. The results show that the proposed model is sensible to phonetic changes, with a 91% of score drops when replacing 20% of the phonemes at random, while providing substantial robustness against different kinds of noise, with a 10% performance drop when mixing the audio with 75% of Gaussian noise. We also provide empirical evidence showing that the resulting embeddings are useful for a variety of downstream applications, such as intelligibility evaluation and the ability to leverage rich pre-trained phonetic embeddings in speech generation task. Finally, we discuss potential applications with interesting implications for the speech generation and recognition fields.

Soil near-Infrared (NIR) spectral absorbance/reflectance libraries are utilized towards improving agricultural production and analysis of soil properties which are key prerequisite for agro-ecological balance and environmental sustainability. Carbonates in particular, represent a soil property which is mostly affected even by mild, let alone extreme, changes of environmental conditions during climate change. In this study we propose a rapid and efficient way to predict carbonates content in soil by means of Fourier Transform Near-Infrared (FT-NIR) reflectance spectroscopy and by use of deep learning methods. We exploited multiple machine learning methods, such as: 1) a Multi-Layered Perceptron Regressor (MLP) and 2) a Convolutional Neural Network (CNN) and compare their performance with other traditional machine learning algorithms such as Partial Least Squares Regression (PLSR), Cubist and Support Vector Machines (SVM) on the combined dataset of two NIR spectral libraries: Kellogg Soil Survey Laboratory (KSSL) of the United States Department of Agriculture (USDA), a dataset of soil samples reflectance spectra collected nationwide, and Land Use and Coverage Area Frame Survey (LUCAS) TopSoil (European Soil Library) which contains soil sample absorbance spectra from all over the European Union, and use them to predict carbonate content on never-before-seen soil samples. Soil samples in KSSL and in TopSoil spectral libraries were acquired in the spectral region of visible-near infrared (Vis-NIR) (350-2500 nm), however in this study, only the NIR spectral region (1150-2500 nm) was utilized. Quantification of carbonates by means of X-ray-Diffraction is in good agreement with the volumetric method and the MLP prediction. Our work contributes to rapid carbonates content prediction in soil samples in cases where: 1) no volumetric method is available and 2) only NIR spectra absorbance data are available. Up till now and to the best of our knowledge, there exists no other study, that presents a prediction model trained on such an extensive dataset with such promising results on unseen data, undoubtedly supporting the notion that deep learning models present excellent prediction tools for soil carbonates content.

This paper introduces Local Learner (2L), an algorithm for providing a set of reference strategies to guide the search for programmatic strategies in two-player zero-sum games. Previous learning algorithms, such as Iterated Best Response (IBR), Fictitious Play (FP), and Double-Oracle (DO), can be computationally expensive or miss important information for guiding search algorithms. 2L actively selects a set of reference strategies to improve the search signal. We empirically demonstrate the advantages of our approach while guiding a local search algorithm for synthesizing strategies in three games, including MicroRTS, a challenging real-time strategy game. Results show that 2L learns reference strategies that provide a stronger search signal than IBR, FP, and DO. We also simulate a tournament of MicroRTS, where a synthesizer using 2L outperformed the winners of the two latest MicroRTS competitions, which were programmatic strategies written by human programmers.

Recent work has studied text-to-audio synthesis using large amounts of paired text-audio data. However, audio recordings with high-quality text annotations can be difficult to acquire. In this work, we approach text-to-audio synthesis using unlabeled videos and pretrained language-vision models. We propose to learn the desired text-audio correspondence by leveraging the visual modality as a bridge. We train a conditional diffusion model to generate the audio track of a video, given a video frame encoded by a pretrained contrastive language-image pretraining (CLIP) model. At test time, we first explore performing a zero-shot modality transfer and condition the diffusion model with a CLIP-encoded text query. However, we observe a noticeable performance drop with respect to image queries. To close this gap, we further adopt a pretrained diffusion prior model to generate a CLIP image embedding given a CLIP text embedding. Our results show the effectiveness of the proposed method, and that the pretrained diffusion prior can reduce the modality transfer gap. While we focus on text-to-audio synthesis, the proposed model can also generate audio from image queries, and it shows competitive performance against a state-of-the-art image-to-audio synthesis model in a subjective listening test. This study offers a new direction of approaching text-to-audio synthesis that leverages the naturally-occurring audio-visual correspondence in videos and the power of pretrained language-vision models.