Nearness of Neighbors Attention for Regression in Supervised Finetuning

It is common in supervised machine learning to combine the feature extraction capabilities of neural networks with the predictive power of traditional algorithms, such as k-nearest neighbors (k-NN) or support vector machines. This procedure involves performing supervised fine-tuning (SFT) on a domain-appropriate feature extractor, followed by training a traditional predictor on the resulting SFT embeddings. When used in this manner, traditional predictors often deliver increased performance over the SFT model itself, despite the fine-tuned feature extractor yielding embeddings specifically optimized for prediction by the neural network's final dense layer. This suggests that directly incorporating traditional algorithms into SFT as prediction layers may further improve performance. However, many traditional algorithms have not been implemented as neural network layers due to their non-differentiable nature and their unique optimization requirements. As a step towards solving this problem, we introduce the Nearness of Neighbors Attention (NONA) regression layer. NONA uses the mechanics of neural network attention and a novel learned attention-masking scheme to yield a differentiable proxy of the k-NN regression algorithm. Results on multiple unstructured datasets show improved performance over both dense layer prediction and k-NN on SFT embeddings for regression.

Longitudinal Ensemble Integration for sequential classification with multimodal data

Effectively modeling multimodal longitudinal data is a pressing need in various application areas, especially biomedicine. Despite this, few approaches exist in the literature for this problem, with most not adequately taking into account the multimodality of the data. In this study, we developed multiple configurations of a novel multimodal and longitudinal learning framework, Longitudinal Ensemble Integration (LEI), for sequential classification. We evaluated LEI's performance, and compared it against existing approaches, for the early detection of dementia, which is among the most studied multimodal sequential classification tasks. LEI outperformed these approaches due to its use of intermediate base predictions arising from the individual data modalities, which enabled their better integration over time. LEI's design also enabled the identification of features that were consistently important across time for the effective prediction of dementia-related diagnoses. Overall, our work demonstrates the potential of LEI for sequential classification from longitudinal multimodal data.

Theoretically informed selection of latent activation in autoencoder based recommender systems

Autoencoders may lend themselves to the design of more accurate and computationally efficient recommender systems by distilling sparse high-dimensional data into dense lower-dimensional latent representations. However, designing these systems remains challenging due to the lack of theoretical guidance. This work addresses this by identifying three key mathematical properties that the encoder in an autoencoder should exhibit to improve recommendation accuracy: (1) dimensionality reduction, (2) preservation of similarity ordering in dot product comparisons, and (3) preservation of non-zero vectors. Through theoretical analysis, we demonstrate that common activation functions, such as ReLU and tanh, cannot fulfill these properties jointly within a generalizable framework. In contrast, sigmoid-like activations emerge as suitable choices for latent activations. This theoretically informed approach offers a more systematic method for hyperparameter selection, enhancing the efficiency of model design.