We show that in-context learning emerges in transformers in discrete developmental stages, when they are trained on either language modeling or linear regression tasks. We introduce two methods for detecting the milestones that separate these stages, by probing the geometry of the population loss in both parameter space and function space. We study the stages revealed by these new methods using a range of behavioral and structural metrics to establish their validity.

We consider the finite-sum optimization problem, where each component function is strongly convex and has Lipschitz continuous gradient and Hessian. The recently proposed incremental quasi-Newton method is based on BFGS update and achieves a local superlinear convergence rate that is dependent on the condition number of the problem. This paper proposes a more efficient quasi-Newton method by incorporating the symmetric rank-1 update into the incremental framework, which results in the condition-number-free local superlinear convergence rate. Furthermore, we can boost our method by applying the block update on the Hessian approximation, which leads to an even faster local convergence rate. The numerical experiments show the proposed methods significantly outperform the baseline methods.

Binary classification, multilabel classification, and regression prediction constitute fundamental paradigms in machine learning, addressing distinct types of predictive modeling tasks. Binary classification involves categorizing instances into one of two classes, typically denoted as positive and negative [1][2][3]. This modeling framework is particularly applicable to scenarios where outcomes are binary in nature, as observed in domains such as spam detection and medical diagnosis. In multilabel classification, the scope extends to situations where instances can be associated with multiple classes simultaneously, a common occurrence in applications like image tagging and document categorization [1][4]. Conversely, regression prediction is concerned with forecasting continuous outcomes, aiming to predict numeric values [3].

Applications involving distributions defined on a hypersphere are remarkably diverse, highlighting the importance of spherical geometries across various disciplines. These applications include: 1) mapping the distribution of geographic or geological features on celestial bodies, such as stars and planets [39, 8, 60], 2) magnetoencephalography (MEG) imaging [75] in medical domains, 3) spherical image representations and 360 images [13, 38], such as omnidirectional images in computer vision [40], 4) texture mapping in computer graphics [24, 21], and more recently, 5) deep representation learning, where the latent representation is often mapped to a bounded space, commonly a sphere, where cosine similarity is utilized for effective representation learning [11, 76]. The analysis of distributions on hyperspheres is traditionally approached through directional statistics, also referred to as circular/spherical statistics [37, 52, 50, 61]. This specialized field is dedicated to the statistical analysis of directions, orientations, and rotations. More recently, with the growing application of optimal transport theory [74, 62] in machine learning, due in part to its favorable statistical, geometrical, and topological properties, there has been an increasing interest in using optimal transport to compare spherical probability measures [14, 32]. One of the main bottlenecks in optimal transport theory is its high computational cost, generally of cubic complexity.

Supervised fairness-aware machine learning under distribution shifts is an emerging field that addresses the challenge of maintaining equitable and unbiased predictions when faced with changes in data distributions from source to target domains. In real-world applications, machine learning models are often trained on a specific dataset but deployed in environments where the data distribution may shift over time due to various factors. This shift can lead to unfair predictions, disproportionately affecting certain groups characterized by sensitive attributes, such as race and gender. In this survey, we provide a summary of various types of distribution shifts and comprehensively investigate existing methods based on these shifts, highlighting six commonly used approaches in the literature. Additionally, this survey lists publicly available datasets and evaluation metrics for empirical studies. We further explore the interconnection with related research fields, discuss the significant challenges, and identify potential directions for future studies.

Recent advances in Graph Neural Networks (GNN) have led to a considerable growth in graph data modeling for multi-modal data which contains various types of nodes and edges. Although some integrative prediction solutions have been developed recently for network-structured data, these methods have some restrictions. For a node classification task involving multi-modal data, certain data modalities may perform better when predicting one class, while others might excel in predicting a different class. Thus, to obtain a better learning representation, advanced computational methodologies are required for the integrative analysis of multi-modal data. Moreover, existing integrative tools lack a comprehensive and cohesive understanding of the rationale behind their specific predictions, making them unsuitable for enhancing model interpretability. Addressing these restrictions, we introduce a novel integrative neural network approach for multi-modal data networks, named Integrative Graph Convolutional Networks (IGCN). IGCN learns node embeddings from multiple topologies and fuses the multiple node embeddings into a weighted form by assigning attention coefficients to the node embeddings. Our proposed attention mechanism helps identify which types of data receive more emphasis for each sample to predict a certain class. Therefore, IGCN has the potential to unravel previously unknown characteristics within different node classification tasks. We benchmarked IGCN on several datasets from different domains, including a multi-omics dataset to predict cancer subtypes and a multi-modal clinical dataset to predict the progression of Alzheimer's disease. Experimental results show that IGCN outperforms or is on par with the state-of-the-art and baseline methods.

It is well known that neuronal systems, including their dynamics and responses, are rife with noise at multiple levels (Faisal et al., 2008; Shadlen & Newsome, 1998). These sources of noise arise from, amongst other things, stochastic processes occuring at the sub-cellular level, impacting neuronal response through, for example, fluctuations in membrane-potential (Derksen & Verveen, 1966). Yet the precise role of such randomness, in information processing, continues to be an open question (McDonnell & Ward, 2011; Deco et al., 2013). The Langevin PC algorithm suggests one such role may be in the principled exploration of the latent space of hypotheses under one's generative model. Secondly, from the perspective of Langevin PC as an in-silico generative modelling algorithm we note a number of interesting avenues that we have not had the time to explore here. These include: Models with a hierarchy of stochastic variables, such as those found in most state of the art VAE models (Child, 2021; Vahdat & Kautz, 2021; Hazami et al., 2022). Which may require adopting a corresponding top-down hierarchical warm-start model. Automatic convergence criteria for determining when our Markov chain has converged to a certain level of error (Roy, 2020). Underdamped Langevin dynamics, which incorporate auxiliary momentum variables into the Langevin sampling to achieve an accelerated rate of convergence (Cheng et al., 2018; Ma et al., 2019).

Text embedding models have emerged as powerful tools for transforming sentences into fixed-sized feature vectors that encapsulate semantic information. While these models are essential for tasks like information retrieval, semantic clustering, and text re-ranking, most existing open-source models, especially those built on architectures like BERT, struggle to represent lengthy documents and often resort to truncation. One common approach to mitigate this challenge involves splitting documents into smaller paragraphs for embedding. However, this strategy results in a much larger set of vectors, consequently leading to increased memory consumption and computationally intensive vector searches with elevated latency. To address these challenges, we introduce Jina Embeddings 2, an open-source text embedding model capable of accommodating up to 8192 tokens. This model is designed to transcend the conventional 512-token limit and adeptly process long documents. Jina Embeddings 2 not only achieves state-of-the-art performance on a range of embedding-related tasks in the MTEB benchmark but also matches the performance of OpenAI's proprietary ada-002 model. Additionally, our experiments indicate that an extended context can enhance performance in tasks such as NarrativeQA.

Speaker diarization has gained considerable attention within speech processing research community. Mainstream speaker diarization rely primarily on speakers' voice characteristics extracted from acoustic signals and often overlook the potential of semantic information. Considering the fact that speech signals can efficiently convey the content of a speech, it is of our interest to fully exploit these semantic cues utilizing language models. In this work we propose a novel approach to effectively leverage semantic information in clustering-based speaker diarization systems. Firstly, we introduce spoken language understanding modules to extract speaker-related semantic information and utilize these information to construct pairwise constraints. Secondly, we present a novel framework to integrate these constraints into the speaker diarization pipeline, enhancing the performance of the entire system. Extensive experiments conducted on the public dataset demonstrate the consistent superiority of our proposed approach over acoustic-only speaker diarization systems.

Machine learning methods for satellite data have a range of societally relevant applications, but labels used to train models can be difficult or impossible to acquire. Self-supervision is a natural solution in settings with limited labeled data, but current self-supervised models for satellite data fail to take advantage of the characteristics of that data, including the temporal dimension (which is critical for many applications, such as monitoring crop growth) and availability of data from many complementary sensors (which can significantly improve a model's predictive performance). We present Presto (the Pretrained Remote Sensing Transformer), a model pre-trained on remote sensing pixel-timeseries data. By designing Presto specifically for remote sensing data, we can create a significantly smaller but performant model. Presto excels at a wide variety of globally distributed remote sensing tasks and performs competitively with much larger models while requiring far less compute. Presto can be used for transfer learning or as a feature extractor for simple models, enabling efficient deployment at scale.