Machine Learning and Data Classification

Comparative Analysis of Machine Learning Libraries for Neural Networks: A Benchmarking Study

This study benchmarks three machine learning libraries on two datasets, finding PyTorch excels in classification tasks while Scikit-learn performs better on simpler tasks.

Leveraging Permutation Testing to Assess Confidence in Positive-Unlabeled Learning Applied to High-Dimensional Biological Datasets

This paper introduces a permutation testing approach to evaluate PU learning models using KP samples without ground truth labels, showing its effectiveness on synthetic and real datasets.

Machine Learning Made Easy (MLme): A Comprehensive Toolkit for Machine Learning-Driven Data Analysis

MLme simplifies machine learning for researchers by integrating data exploration, AutoML, and visualization, showing promising performance across datasets and identifying cell population markers.

DataSAIL: Data Splitting Against Information Leakage

DataSAIL splits biological datasets to minimize information leakage by reducing similarity between splits, improving model generalization assessment.

Paired evaluation defines performance landscapes for machine learning models

This study introduces paired evaluation to improve performance estimation of machine learning models in drug response and disease severity prediction, showing up to 30% variation in performance estimates based on test data choice.

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.

Correcting machine learning models using calibrated ensembles with "mlensemble"

This paper introduces mlensemble, a software package that corrects machine learning models by calibrating ensembles to new data, improving performance across different datasets and applications.

Just Add Data: Automated Predictive Modeling and BioSignature Discovery

The paper introduces JADBIO, an AutoML system for omics data that produces predictive models and biosignatures, showing it can overcome overfitting in high-dimensional, low-sample data.

I TRIED A BUNCH OF THINGS: THE DANGERS OF UNEXPECTED OVERFITTING IN CLASSIFICATION

The paper demonstrates how overfitting through hyperparameter tuning can lead to non-generalizable results in machine learning, despite cross-validation, and suggests safeguards to prevent this.

Tree-weighting for multi-study ensemble learners

This paper develops tree-weighting methods for multi-study ensemble learning, showing improved robustness and feature interpretation in genomic datasets.

Scaling tree-based automated machine learning to biomedical big data with a dataset selector

The paper introduces two features to TPOT, a machine learning tool, to improve its scalability and efficiency for biomedical big data analysis.