We show that variational learning can significantly improve the accuracy and calibration of Low-Rank Adaptation (LoRA) without a substantial increase in the cost. We replace AdamW by the Improved Variational Online Newton (IVON) algorithm to finetune large language models. For Llama-2 with 7 billion parameters, IVON improves the accuracy over AdamW by 2.8% and expected calibration error by 4.6%. The accuracy is also better than the other Bayesian alternatives, yet the cost is lower and the implementation is easier. Our work provides additional evidence for the effectiveness of IVON for large language models.

Given a database of bit strings $A_1,\ldots,A_m\in \{0,1\}^n$, a fundamental data structure task is to estimate the distances between a given query $B\in \{0,1\}^n$ with all the strings in the database. In addition, one might further want to ensure the integrity of the database by releasing these distance statistics in a secure manner. In this work, we propose differentially private (DP) data structures for this type of tasks, with a focus on Hamming and edit distance. On top of the strong privacy guarantees, our data structures are also time- and space-efficient. In particular, our data structure is $\epsilon$-DP against any sequence of queries of arbitrary length, and for any query $B$ such that the maximum distance to any string in the database is at most $k$, we output $m$ distance estimates. Moreover, - For Hamming distance, our data structure answers any query in $\widetilde O(mk+n)$ time and each estimate deviates from the true distance by at most $\widetilde O(k/e^{\epsilon/\log k})$; - For edit distance, our data structure answers any query in $\widetilde O(mk^2+n)$ time and each estimate deviates from the true distance by at most $\widetilde O(k/e^{\epsilon/(\log k \log n)})$. For moderate $k$, both data structures support sublinear query operations. We obtain these results via a novel adaptation of the randomized response technique as a bit flipping procedure, applied to the sketched strings.

Bayes' rule naturally allows for inference refinement in a streaming fashion, without the need to recompute posteriors from scratch whenever new data arrives. In principle, Bayesian streaming is straightforward: we update our prior with the available data and use the resulting posterior as a prior when processing the next data chunk. In practice, however, this recipe entails i) approximating an intractable posterior at each time step; and ii) encapsulating results appropriately to allow for posterior propagation. For continuous state spaces, variational inference (VI) is particularly convenient due to its scalability and the tractability of variational posteriors. For discrete state spaces, however, state-of-the-art VI results in analytically intractable approximations that are ill-suited for streaming settings. To enable streaming Bayesian inference over discrete parameter spaces, we propose streaming Bayes GFlowNets (abbreviated as SB-GFlowNets) by leveraging the recently proposed GFlowNets -- a powerful class of amortized samplers for discrete compositional objects. Notably, SB-GFlowNet approximates the initial posterior using a standard GFlowNet and subsequently updates it using a tailored procedure that requires only the newly observed data. Our case studies in linear preference learning and phylogenetic inference showcase the effectiveness of SB-GFlowNets in sampling from an unnormalized posterior in a streaming setting. As expected, we also observe that SB-GFlowNets is significantly faster than repeatedly training a GFlowNet from scratch to sample from the full posterior.

Mixtures of Experts (MoE) are Machine Learning models that involve partitioning the input space, with a separate "expert" model trained on each partition. Recently, MoE have become popular as components in today's large language models as a means to reduce training and inference costs. There, the partitioning function and the experts are both learnt jointly via gradient descent on the log-likelihood. In this paper we focus on studying the efficiency of the Expectation Maximization (EM) algorithm for the training of MoE models. We first rigorously analyze EM for the cases of linear or logistic experts, where we show that EM is equivalent to Mirror Descent with unit step size and a Kullback-Leibler Divergence regularizer. This perspective allows us to derive new convergence results and identify conditions for local linear convergence based on the signal-to-noise ratio (SNR). Experiments on synthetic and (small-scale) real-world data show that EM outperforms the gradient descent algorithm both in terms of convergence rate and the achieved accuracy.

Many important phenomena in scientific fields such as climate, neuroscience, and epidemiology are naturally represented as spatiotemporal gridded data with complex interactions. For example, in climate science, researchers aim to uncover how large-scale events, such as the North Atlantic Oscillation (NAO) and the Antarctic Oscillation (AAO), influence other global processes. Inferring causal relationships from these data is a challenging problem compounded by the high dimensionality of such data and the correlations between spatially proximate points. We present SPACY (SPAtiotemporal Causal discoverY), a novel framework based on variational inference, designed to explicitly model latent time-series and their causal relationships from spatially confined modes in the data. Our method uses an end-to-end training process that maximizes an evidence-lower bound (ELBO) for the data likelihood. Theoretically, we show that, under some conditions, the latent variables are identifiable up to transformation by an invertible matrix. Empirically, we show that SPACY outperforms state-of-the-art baselines on synthetic data, remains scalable for large grids, and identifies key known phenomena from real-world climate data.

Missing data is a common problem in finance and often requires methods to fill in the gaps, or in other words, imputation. In this work, we focused on the imputation of missing implied volatilities for FX options. Prior work has used variational autoencoders (VAEs), a neural network-based approach, to solve this problem; however, using stronger classical baselines such as Heston with jumps can significantly outperform their results. We show that simple modifications to the architecture of the VAE lead to significant imputation performance improvements (e.g., in low missingness regimes, nearly cutting the error by half), removing the necessity of using $\beta$-VAEs. Further, we modify the VAE imputation algorithm in order to better handle the uncertainty in data, as well as to obtain accurate uncertainty estimates around imputed values.

Data compression techniques that are focused on information preservation have become essential in the modern era of big data. In this work, an encoderdecoder architecture has been designed where adversaria! The goal of this proposal is to obtain an intermediate representation of the astronomical stellar spectra, in which the contribution to the flux of a star due to the most influential physical properties (its surface temperature and gravity) disappears and the variance reflects only the effect of the chemical composition over the spectrum. We apply a scheme of deep learning with the aim of unraveling in the latent space the desired parameters of the rest of the information contained in the data. This work propases a version of adversaria! training that makes use of one discriminator per parameter to be disentangled, thus avoiding the exponential combination that occurs in the use of a single discriminator, as a result of the discretization of the values to be untangled. To test the effectiveness of the method, synthetic astronomical data are used from the APOGEE and Gaia surveys. In conjunction with the work presented, an ad-hoc framework (GANDALF) is provided, which allows the replication, visualization, and extension of the method to domains of any nature. Keywords: Generative Adversaria! Neural Networks, Disentangled Representation, Astronomical Spectra, Gaia mission, APOGEE Preprint submitted to Applied Soft Computing November 8, 2024 l. Introduction Finding representations of the data that can ease the extraction of useful inforrnation and irnprove algorithrn performance in classification or pararnetrization problerns has becorne a field in itself in the rnachine learning cornrnunity, and is known as representation learning. The process of unraveling these underlying factors in a cornprehensive representation is called disentangled representation. There is abundant literature on the problern of how to decode or separate representations of a signal into projections that include only inforrnation relevant to a specific problern. In Wang et al. [1] the current state of the literature is exhaustively reviewed discussing different rnethodologies, rnetrics, rnodels, and applications.

Dose prediction is an area of ongoing research that facilitates radiotherapy planning. Most commercial models utilise imaging data and intense computing resources. This study aimed to predict the dose-volume of rectum and bladder from volumes of target, at-risk structure organs and their overlap regions using machine learning. Dose-volume information of 94 patients with prostate cancer planned for 6000cGy in 20 fractions was exported from the treatment planning system as text files and mined to create a training dataset. Several statistical modelling, machine learning methods, and a new fuzzy rule-based prediction (FRBP) model were explored and validated on an independent dataset of 39 patients. The median absolute error was 2.0%-3.7% for bladder and 1.7-2.4% for rectum in the 4000-6420cGy range. For 5300cGy, 5600cGy and 6000cGy, the median difference was less than 2.5% for rectum and 3.8% for bladder. The FRBP model produced errors of 1.2%, 1.3%, 0.9% and 1.6%, 1.2%, 0.1% for the rectum and bladder respectively at these dose levels. These findings indicate feasibility of obtaining accurate predictions of the clinically important dose-volume parameters for rectum and bladder using just the volumes of these structures.

Autoregressive modeling has been a huge success in the field of natural language processing (NLP). Recently, autoregressive models have emerged as a significant area of focus in computer vision, where they excel in producing high-quality visual content. Autoregressive models in NLP typically operate on subword tokens. However, the representation strategy in computer vision can vary in different levels, \textit{i.e.}, pixel-level, token-level, or scale-level, reflecting the diverse and hierarchical nature of visual data compared to the sequential structure of language. This survey comprehensively examines the literature on autoregressive models applied to vision. To improve readability for researchers from diverse research backgrounds, we start with preliminary sequence representation and modeling in vision. Next, we divide the fundamental frameworks of visual autoregressive models into three general sub-categories, including pixel-based, token-based, and scale-based models based on the strategy of representation. We then explore the interconnections between autoregressive models and other generative models. Furthermore, we present a multi-faceted categorization of autoregressive models in computer vision, including image generation, video generation, 3D generation, and multi-modal generation. We also elaborate on their applications in diverse domains, including emerging domains such as embodied AI and 3D medical AI, with about 250 related references. Finally, we highlight the current challenges to autoregressive models in vision with suggestions about potential research directions. We have also set up a Github repository to organize the papers included in this survey at: \url{https://github.com/ChaofanTao/Autoregressive-Models-in-Vision-Survey}.

The Gaussian process (GP) is a widely used probabilistic machine learning method for stochastic function approximation, stochastic modeling, and analyzing real-world measurements of nonlinear processes. Unlike many other machine learning methods, GPs include an implicit characterization of uncertainty, making them extremely useful across many areas of science, technology, and engineering. Traditional implementations of GPs involve stationary kernels (also termed covariance functions) that limit their flexibility and exact methods for inference that prevent application to data sets with more than about ten thousand points. Modern approaches to address stationarity assumptions generally fail to accommodate large data sets, while all attempts to address scalability focus on approximating the Gaussian likelihood, which can involve subjectivity and lead to inaccuracies. In this work, we explicitly derive an alternative kernel that can discover and encode both sparsity and nonstationarity. We embed the kernel within a fully Bayesian GP model and leverage high-performance computing resources to enable the analysis of massive data sets. We demonstrate the favorable performance of our novel kernel relative to existing exact and approximate GP methods across a variety of synthetic data examples. Furthermore, we conduct space-time prediction based on more than one million measurements of daily maximum temperature and verify that our results outperform state-of-the-art methods in the Earth sciences. More broadly, having access to exact GPs that use ultra-scalable, sparsity-discovering, nonstationary kernels allows GP methods to truly compete with a wide variety of machine learning methods.