A Flow Model with Low-Rank Transformers for Incomplete Multimodal Survival Analysis

DeepBlip is a novel neural framework designed to estimate conditional average treatment effects over time using structural nested mean models (SNMMs). This approach allows for the decomposition of treatment sequences into localized, time-specific 'blip effects', enhancing interpretability and enabling efficient evaluation of treatment policies. DeepBlip integrates sequential neural networks like LSTMs and transformers, addressing the limitations of existing methods by allowing simultaneous learning of all blip functions.

The study explores the scaling laws of deep neural networks in medical anatomical segmentation, revealing that larger training datasets lead to improved performance across various semantic tasks and imaging modalities. It highlights the significance of deformation-guided augmentation strategies, such as random elastic deformation and registration-guided deformation, in enhancing segmentation outcomes. The research aims to address the underexplored area of data scaling in medical imaging, proposing a novel image augmentation approach to generate diffeomorphic mappings.

The paper introduces the attention-indexed model (AIM), a framework for analyzing learning in deep attention layers. AIM captures the emergence of token-level outputs from bilinear interactions over high-dimensional embeddings. It allows full-width key and query matrices, aligning with practical transformers. The study derives predictions for Bayes-optimal generalization error and identifies phase transitions based on sample complexity, model width, and sequence length, proposing a message passing algorithm and demonstrating optimal performance via gradient descent.

Recent experiments indicate that the training trajectories of various deep neural networks, regardless of their architecture or optimization methods, follow a low-dimensional 'hyper-ribbon-like' manifold in probability distribution space. This study analytically characterizes this behavior in linear networks, revealing that the manifold's geometry is influenced by factors such as the decay rate of eigenvalues from the input correlation matrix, the initial weight scale, and the number of gradient descent steps.

CLAReSNet, a new hybrid architecture for hyperspectral image classification, integrates multi-scale convolutional extraction with transformer-style attention through an adaptive latent bottleneck. This model addresses challenges such as high spectral dimensionality, complex spectral-spatial correlations, and limited training samples with severe class imbalance. By combining convolutional networks and transformers, CLAReSNet aims to enhance classification accuracy and efficiency in hyperspectral imaging applications.

Higher-order Neural Additive Models (HONAMs) have been introduced as an advancement over Neural Additive Models (NAMs), which are known for their predictive performance and interpretability. HONAMs address the limitation of NAMs by effectively capturing feature interactions of arbitrary orders, enhancing predictive accuracy while maintaining interpretability, crucial for high-stakes applications. The source code for HONAM is publicly available on GitHub.

Vision-Language Models (VLMs) are emerging models that integrate visual content with natural language. Current methods typically fuse data either early in the encoding process or late through pooled embeddings. This paper introduces a lightweight fusion module utilizing cross-only, bidirectional attention layers to align hidden states from both modalities, enhancing understanding while keeping encoders non-causal. The proposed method aims to improve the performance of VLMs by leveraging the inherent structure of visual and textual data.