Randomized algorithms can be used to speed up the analysis of large datasets. In this paper, we develop a unified methodology for statistical inference via randomized sketching or projections in two of the most fundamental problems in multivariate statistical analysis: least squares and PCA. The methodology applies to fixed datasets -- i.e., is data-conditional -- and the only randomness is due to the randomized algorithm. We propose statistical inference methods for a broad range of sketching distributions, such as the subsampled randomized Hadamard transform (SRHT), Sparse Sign Embeddings (SSE) and CountSketch, sketching matrices with i.i.d. entries, and uniform subsampling. To our knowledge, no comparable methods are available for SSE and for SRHT in PCA. Our novel theoretical approach rests on showing the asymptotic normality of certain quadratic forms. As a contribution of broader interest, we show central limit theorems for quadratic forms of the SRHT, relying on a novel proof via a dyadic expansion that leverages the recursive structure of the Hadamard transform. Numerical experiments using both synthetic and empirical datasets support the efficacy of our methods, and in particular suggest that sketching methods can have better computation-estimation tradeoffs than recently proposed optimal subsampling methods.

By training to predict the next token in an unlabeled corpus, large language models learn to perform many tasks without any labeled data. However, their next-token-prediction objective arguably limits their performance in scenarios that require planning, such as writing a coherent article. In this paper, we train a module for planning the future writing process via a self-supervised learning objective. By conditioning on generated latent plans, our model extends the successful language model formula to more abstract planning in an unsupervised way. Empirically, we demonstrate that our method improves language modeling performance in general, particularly with respect to the text structure. Because our framework uses a planner module that is unsupervised and external to the language model, new planner modules can be trained at large scale and easily be shared with the community.

Social media data has been used for detecting users with mental disorders, such as depression. Despite the global significance of cross-cultural representation and its potential impact on model performance, publicly available datasets often lack crucial metadata related to this aspect. In this work, we evaluate the generalization of benchmark datasets to build AI models on cross-cultural Twitter data. We gather a custom geo-located Twitter dataset of depressed users from seven countries as a test dataset. Our results show that depression detection models do not generalize globally. The models perform worse on Global South users compared to Global North. Pre-trained language models achieve the best generalization compared to Logistic Regression, though still show significant gaps in performance on depressed and non-Western users. We quantify our findings and provide several actionable suggestions to mitigate this issue.

This paper introduces SOAR: Spilling with Orthogonality-Amplified Residuals, a novel data indexing technique for approximate nearest neighbor (ANN) search. SOAR extends upon previous approaches to ANN search, such as spill trees, that utilize multiple redundant representations while partitioning the data to reduce the probability of missing a nearest neighbor during search. Rather than training and computing these redundant representations independently, however, SOAR uses an orthogonality-amplified residual loss, which optimizes each representation to compensate for cases where other representations perform poorly. This drastically improves the overall index quality, resulting in state-of-the-art ANN benchmark performance while maintaining fast indexing times and low memory consumption.

Abstract--Machine learning (ML) methods are widely used most informative data points benefiting the downstream tasks in manufacturing applications, which usually require a large and mitigate the impact of low-quality data. We collected realworld amount of training data. However, data collection needs extensive in-situ monitoring data of the same additive manufacturing costs and time investments in the manufacturing system, and process from three different machines, two of which are more data scarcity commonly exists. The proposed method is applied to train an industrial internet of things (IIoT), data-sharing is widely enabled anomaly detection model for those two similar machines, and the among multiple machines with similar functionality to augment entire data pool from all three machines is available for selecting the dataset for building ML models. The results demonstrated that our proposed designed similarly, the distribution mismatch inevitably exists in method outperforms the benchmark methods by only requiring their data due to different working conditions, process parameters, 26% of labeled training samples. In addition, all selected data measurement noise, etc. However, the effective application samples are from machines with similar conditions, while the of ML methods is built upon the assumption that the training data from the different machines are prevented from misleading and testing data are sampled from the same distribution. In this work, we propose an Active Data-sharing (ADs) framework to ensure the quality of the shared data among multiple machines. Low-quality data here refers to data samples collected from machines/processes different from the target one.

Rhetorical Role Labeling (RRL) of legal judgments is essential for various tasks, such as case summarization, semantic search and argument mining. However, it presents challenges such as inferring sentence roles from context, interrelated roles, limited annotated data, and label imbalance. This study introduces novel techniques to enhance RRL performance by leveraging knowledge from semantically similar instances (neighbours). We explore inference-based and training-based approaches, achieving remarkable improvements in challenging macro-F1 scores. For inference-based methods, we explore interpolation techniques that bolster label predictions without re-training. While in training-based methods, we integrate prototypical learning with our novel discourse-aware contrastive method that work directly on embedding spaces. Additionally, we assess the cross-domain applicability of our methods, demonstrating their effectiveness in transferring knowledge across diverse legal domains.

In this paper, we introduce a nonparametric end-to-end method for probabilistic forecasting of distributed renewable generation outputs while including missing data imputation. Firstly, we employ a nonparametric probabilistic forecast model utilizing the long short-term memory (LSTM) network to model the probability distributions of distributed renewable generations' outputs. Secondly, we design an end-to-end training process that includes missing data imputation through iterative imputation and iterative loss-based training procedures. This two-step modeling approach effectively combines the strengths of the nonparametric method with the end-to-end approach. Consequently, our approach demonstrates exceptional capabilities in probabilistic forecasting for the outputs of distributed renewable generations while effectively handling missing values. Simulation results confirm the superior performance of our approach compared to existing alternatives.

Do transformers "think ahead" during inference at a given position? It is known transformers prepare information in the hidden states of the forward pass at $t$ that is then used in future forward passes $t+\tau$. We posit two explanations for this phenomenon: pre-caching, in which off-diagonal gradient terms present in training result in the model computing features at $t$ irrelevant to the present inference task but useful for the future, and breadcrumbs, in which features most relevant to time step $t$ are already the same as those that would most benefit inference at time $t+\tau$. We test these hypotheses by training language models without propagating gradients to past timesteps, a scheme we formalize as myopic training. In a synthetic data setting, we find clear evidence for pre-caching. In the autoregressive language modeling setting, our experiments are more suggestive of the breadcrumbs hypothesis.

Wikipedia articles (content pages) are commonly used corpora in Natural Language Processing (NLP) research, especially in low-resource languages other than English. Yet, a few research studies have studied the three Arabic Wikipedia editions, Arabic Wikipedia (AR), Egyptian Arabic Wikipedia (ARZ), and Moroccan Arabic Wikipedia (ARY), and documented issues in the Egyptian Arabic Wikipedia edition regarding the massive automatic creation of its articles using template-based translation from English to Arabic without human involvement, overwhelming the Egyptian Arabic Wikipedia with articles that do not only have low-quality content but also with articles that do not represent the Egyptian people, their culture, and their dialect. In this paper, we aim to mitigate the problem of template translation that occurred in the Egyptian Arabic Wikipedia by identifying these template-translated articles and their characteristics through exploratory analysis and building automatic detection systems. We first explore the content of the three Arabic Wikipedia editions in terms of density, quality, and human contributions and utilize the resulting insights to build multivariate machine learning classifiers leveraging articles' metadata to detect the template-translated articles automatically. We then publicly deploy and host the best-performing classifier, XGBoost, as an online application called EGYPTIAN WIKIPEDIA SCANNER and release the extracted, filtered, and labeled datasets to the research community to benefit from our datasets and the online, web-based detection system.

The clustering method based on the anchor graph has gained significant attention due to its exceptional clustering performance and ability to process large-scale data. One common approach is to learn bipartite graphs with K-connected components, helping avoid the need for post-processing. However, this method has strict parameter requirements and may not always get K-connected components. To address this issue, an alternative approach is to directly obtain the cluster label matrix by performing non-negative matrix factorization (NMF) on the anchor graph. Nevertheless, existing multi-view clustering methods based on anchor graph factorization lack adequate cluster interpretability for the decomposed matrix and often overlook the inter-view information. We address this limitation by using non-negative tensor factorization to decompose an anchor graph tensor that combines anchor graphs from multiple views. This approach allows us to consider inter-view information comprehensively. The decomposed tensors, namely the sample indicator tensor and the anchor indicator tensor, enhance the interpretability of the factorization. Extensive experiments validate the effectiveness of this method.