contrastive learning

Dirac mixture distributions for the approximation of mixed effects models

The paper compares methods for approximating mixed effects models, finding that the modified Cramer-von Mises Distance method provides better accuracy than Monte-Carlo and sigma-point methods.

Cellular structure image classification with small targeted training samples

The paper presents a deep learning method using GANs and feature transfer to classify cell shapes with minimal training samples, achieving high accuracy in detecting multicellular rosettes in C. elegans embryos.

Predicting the Evolution of White Matter Hyperintensities in Brain MRI using Generative Adversarial Networks and Irregularity Map

This paper introduces DEP-GAN, a GAN model that predicts WMH evolution in small vessel disease using irregularity maps and two discriminators, achieving 72.01% accuracy in predicting WMH evolution.

Linear-Nonlinear-Time-Warp-Poisson models of neural activity

The paper introduces Linear-Nonlinear-Time-Warp-Poisson models to account for shared temporal variability in neural activity, showing improved predictions in macaque premotor cortex recordings.

Cryptography for genetic material

This paper describes a cryptographic method to conceal genetic information using random barcode subsets and pooling, evaluated with differential privacy.

ENRICH: a fast method to improve the quality of flexible macromolecular reconstructions

ENRICH improves 3D map quality from small, heterogeneous cryo-EM datasets by merging particles across conformations using optical flow.

Fast Estimation of Recombination Rates Using Topological Data Analysis

This paper introduces TREE, a method using topological data analysis to estimate recombination rates more accurately and efficiently than previous methods.

Performance of convolutional neural networks for identification of bacteria in 3D microscopy datasets

This study evaluates 3D convolutional neural networks for identifying bacteria in zebrafish intestinal microscopy images, finding they match human accuracy and outperform traditional methods.

BindSpace: decoding transcription factor binding signals by large-scale joint embedding

BindSpace is a model that embeds DNA sequences and TF labels into a shared space, achieving state-of-the-art binding prediction for hundreds of TFs.

Attend and Predict: Understanding Gene Regulation by Selective Attention on Chromatin

AttentiveChrome combines hierarchical Long Short-Term Memory networks with two-level attention to predict human gene expression from chromatin marks across 56 cell types while identifying important marks and genomic positions.

Decoding of neural data using cohomological learning

This paper introduces cohomological learning to decode neural features from mouse movement data, identifying head direction cells and finding that decoded values contain more information than tracked head direction.

Generative adversarial networks as a tool to recover structural information from cryo-electron microscopy data

This paper uses generative adversarial networks to denoise cryo-EM data, improving structural information recovery from noisy images.

Insular Celtic population structure and genomic footprints of migration

This paper uses haplotype-based methods to reveal genetic structure in Ireland, showing regional diversity and historical migrations, with implications for genetic association studies.

Trajectory-Based Parameterization of a Coarse-Grained Forcefield for High-Throughput Protein Simulation

The paper trains a coarse-grained forcefield using contrastive divergence on simulation trajectories to improve protein folding accuracy while maintaining computational efficiency.

Applying Mondrian Cross-Conformal Prediction to Estimate Prediction Confidence on Large Imbalanced Bioactivity Datasets

This study applies Mondrian cross-conformal prediction to estimate prediction confidence on large imbalanced bioactivity datasets, showing improved performance for minority classes.

RNA-protein binding motifs mining with a new hybrid deep learning based cross-domain knowledge integration approach

The paper introduces iDeep, a hybrid deep learning framework that improves RNA-protein binding motif prediction by integrating cross-domain knowledge, achieving better performance and interpretability.

Local Discriminant Hyperalignment for multi-subject fMRI data alignment

This paper introduces LDHA, a supervised hyperalignment method for multi-subject fMRI data that improves performance in multivariate pattern analysis by incorporating local discriminant analysis.