contrastive learning

Learning neural decoders without labels using multiple data streams

This paper develops cross-modal, self-supervised neural decoders using multiple data streams to achieve high decoding accuracy without labeled data.

Improving deep learning-based segmentation of diatoms in gigapixel-sized virtual slides by object-based tile positioning and object integrity constraint

This paper improves deep learning segmentation of diatoms in large images by using object-based tiling with integrity constraints, achieving higher precision without sacrificing recall.

Differential geometry methods for constructing manifold-targeted recurrent neural networks.

This paper introduces a method using differential geometry to construct recurrent neural networks with dynamics targeted to a specific manifold, improving the design of task-solving neural dynamics.

ProSelfLC: Progressive Self Label Correction Towards A Low-Temperature Entropy State

ProSelfLC improves DNN robustness by progressively trusting low-entropy predictions and correcting labels during training.

Discovering molecular features of intrinsically disordered regions by using evolution for contrastive learning

The paper introduces a method to discover molecular features of intrinsically disordered regions using contrastive learning with evolutionary conservation.

MB-SupCon: Microbiome-based predictive models via Supervised Contrastive Learning

MB-SupCon integrates microbiome and metabolomics data to improve prediction accuracy in microbiome-only datasets, achieving high accuracy in predicting insulin resistance, sex, and race.

Direct Generation of Protein Conformational Ensembles via Machine Learning

This paper shows that machine learning can generate protein conformational ensembles without simulation, using a transformer-based GAN trained on simulation data.

Selfee: Self-supervised Features Extraction of animal behaviors

Selfee extracts comprehensive and discriminative features from raw animal behavior videos using self-supervised learning, enabling anomaly detection and temporal analysis.

Effective Representation Learning via The Integrated Self-Supervised Pre-training models of StyleGAN2-ADA and DINO for Colonoscopy Images

The paper combines StyleGAN2-ADA and DINO for self-supervised pre-training to learn representations from colonoscopy images without annotated data.

Uncovering features of synapses in primary visual cortex through contrastive representation learning

This paper introduces SynapseCLR, a contrastive learning method for 3D EM data to extract synapse features from mouse visual cortex, achieving high accuracy in synapse classification and annotation.

TANKBind: Trigonometry-Aware Neural NetworKs for Drug-Protein Binding Structure Prediction

TANKBind improves drug-protein binding structure prediction by incorporating trigonometry constraints and functional block segmentation, achieving better performance than existing methods.

Transfer Learning for Cognitive Reserve Quantification

This paper develops a deep learning model to quantify cognitive reserve using structural MRI data from healthy cohorts and validates its generalizability across different populations and scanners using transfer learning.

Transfer learning identifies sequence determinants of regulatory element accessibility

This paper introduces ChromTransfer, a transfer learning method that improves cell-type specific chromatin accessibility prediction from DNA sequences using pre-trained models.

Adversarial domain translation networks for fast and accurate integration of large-scale atlas-level single-cell datasets

Portal integrates large-scale single-cell datasets efficiently and accurately by preserving biological variation across samples, platforms, and species.

Design of Peptide-Based Protein Degraders via Contrastive Deep Learning

This paper designs peptides to degrade undruggable proteins using contrastive deep learning and experimental validation in human cells.

Riemannian geometry of functional connectivity matrices for multi-site attention-deficit/hyperactivity disorder data harmonization

This paper introduces a geometric harmonization method for multi-site functional connectivity data, showing it improves alignment with pathophysiological findings in ADHD studies.

Convolutions are competitive with transformers for protein sequence pretraining

This study shows that convolutional neural networks can be as effective as transformers for protein sequence pretraining, offering better scalability and performance on long sequences.

MorphoGNN: Morphological Embedding for Single Neuron with Graph Neural Networks

The paper introduces MorphoGNN, a graph neural network method for embedding single neuron morphology, achieving competitive performance in classification and retrieval tasks.

ReCSAI Recursive compressed sensing artificial intelligence for confocal lifetime localization microscopy

The paper introduces ReCSAI, a compressed sensing neural network for confocal dSTORM that improves reconstruction speed and accuracy by combining deep learning and compressed sensing.

Content-aware image restoration improves spatiotemporal resolution in luminescence microscopy.

This paper shows that content-aware image restoration reduces exposure time in luminescence microscopy, improving temporal resolution.

An Ensemble Learning and Slice Fusion Strategy for Three-Dimensional Nuclei Instance Segmentation

This paper introduces EMR-CNN, a 3D nuclei segmentation method that avoids ground truth annotations and uses ensemble learning and slice fusion for efficient and accurate results.

Contrastive learning on protein embeddings enlightens midnight zone

This paper introduces ProtTucker, a contrastive learning method using protein embeddings from pLMs to improve recognition of distant homologous relationships in the midnight zone of protein similarity.

Deep learning to decompose macromolecules into independent Markovian domains

This paper introduces a method combining independent Markov decomposition with VAMPnets to decompose large molecular systems into Markovian subdomains, improving data efficiency and interpretability.

Pneumonia Detection with Semantic Similarity Scores

This paper develops an abnormality detection method for pneumonia using self-supervised contrastive learning and Mahalanobis distance scores, outperforming previous unsupervised methods.

Self-Supervised Natural Image Reconstruction and Large-Scale Semantic Classification from Brain Activity

This paper develops a self-supervised method to reconstruct images and classify semantic categories from fMRI data without paired training examples.

Subcortical Segmentation of the Fetal Brain in 3D Ultrasound using Deep Learning

The paper develops a CNN for automated segmentation of fetal brain structures in 3D ultrasound using few-shot learning, achieving performance close to intra-observer variability with minimal manual annotations.

Disentanglement of Entropy and Coevolution using Spectral Regularization

This paper introduces a spectral regularizer to disentangle entropy from coevolution in sequence models, improving interpretability and reducing overfitting in contact map inference.

Class-Controlled Copy-Paste Based Cell Segmentation for CoNIC Challenge

The paper introduces a class-controlled copy-paste augmentation method to address class imbalance in histopathology cell segmentation, improving metrics for underrepresented classes.

The RNA Virome of Echinoderms

This study characterizes the RNA virome of 38 echinoderm species, identifying 347 viral genome fragments across nine viral orders and finding no significant difference in viral reads between life stages.

Deep Augmentation for Electrode Shift Compensation in Transient High-density sEMG: Towards Application in Neurorobotics

This paper introduces deep data augmentation for transient HD-sEMG to improve gesture prediction and robustness against electrode shift in neurorobotics applications.