Current challenges in developing foundational models for volumetric imaging data, such as magnetic resonance imaging (MRI), stem from the computational complexity of state-of-the-art architectures in high dimensions and curating sufficiently large datasets of volumes. To address these challenges, we introduce Raptor (Random Planar Tensor Reduction), a train-free method for generating semantically rich embeddings for volumetric data. Raptor leverages a frozen 2D foundation model, pretrained on natural images, to extract visual tokens from individual cross-sections of medical volumes. These tokens are then spatially compressed using random projections, significantly reducing computational complexity while retaining rich semantic information. Extensive experiments on 10 diverse medical volume tasks verify the superior performance of Raptor over state-of-the-art methods, including those pretrained exclusively on medical volumes (+3% SuPreM, +6% MISFM, +10% Merlin, +13% VoCo, and +14% SLIViT), while entirely bypassing the need for costly training. Our results highlight Raptor’s effectiveness and versatility as a foundation for advancing deep learning-based methods for medical volumes (code: github.com/sriramlab/raptor).

We present CACTI, a masked autoencoding approach for imputing tabular data that leverages the structure in missingness patterns and contextual information. Our approach employs a novel median truncated copy masking training strategy that encourages the model to learn from empirical patterns of missingness while incorporating semantic relationships between features — captured by column names and text descriptions — to better represent feature dependence. These dual sources of inductive bias enable CACTIto outperform state-of-the-art methods — an average $R^2$ gain of 7.8\% over the next best method (13.4%, 6.1%, and 5.3% under missing not at random, at random and completely at random, respectively) — across a diverse range of datasets and missingness conditions. Our results highlight the value of leveraging dataset-specific contextual information and missingness patterns to enhance imputation performance.

The ability to forecast traffic congestion ahead of time given road conditions has remained a prominent problem in road traffic analysis. In this work, we leverage mobility traces of public transport vehicles tracked by the New York City MTA and formulate Message-Passing Recurrent Neural Nets (MPRNN) to produce long-term traffic forecasting on data that is sparse but wide in coverage. We model the interactions among road segments spread over the entirety of Manhattan, New York over a period of 3 months, such that traffic conditions can be propagated to > 90% of examined segments from just a few observations. In comparison to other competing algorithms, MPRNN achieves the lowest mean error of <; 0.3 mph when predicting ahead in 10 minute intervals, for up to 3 road segments ahead (message passing across 3 hops). The MPRNN model further offers compelling results when forecasting traffic speeds several hours ahead given distant observations up to approximately 1 kilometer away (three consecutive bus stops) with a mean error of about 2 mph.