Machine Learning for Research

Explore the steps of the scientific process from formulating a research question to publishing findings, and see how machine learning enhances data collection and analysis.

Unlock automation and efficiency in data collection and preprocessing, reducing bias and freeing researchers for analysis. Apply machine learning to high-dimensional, multimodal data for personalized medicine and predictive modeling.

Machine learning in action shows how algorithms diagnose eye diseases like AMD with high accuracy and enable earlier treatment. It shows medicine's future with accessible, fast machine learning tools.

Keep researchers in the driver's seat by evaluating when ML is necessary, ensuring data quality and representativeness, and maintaining interpretability and a clear research design.

Explore how machine learning fits scientific process, aiding data collection and analysis of complex patterns. Handle multimodal data, assess when to apply it, ensure data quality, and consider solutions first.

Learn to preprocess data for machine learning, addressing missing values, outliers, and errors to improve data quality. Explore data augmentation, SMOTE, and PCA, and training-test split in ML workflow.

Explore data augmentation, a simple method creating new data by modifying samples. Expand datasets and curb overfitting with bias-free data; follow the rule of ten, illustrated by colorectal cancer histology.

Demonstrates data augmentation with TensorFlow for histology slides, including rotation 90 degrees, center cropping, and Gaussian noise to boost dataset size and model robustness.

Explore how Smote addresses class imbalance by generating synthetic minority data via nearest-neighbor interpolation using Euclidean distance, typically with k=5, to reach a 1:1 ratio.

Demonstrates implementing smote to balance an imbalanced breast cancer dataset by oversampling the malignant class, then visualizes feature-space changes before and after.

Explore principal components analysis (PCA) and its use in reducing dimensions for research, creating principal components and loadings that reveal key features and support visualization, feature selection, and generalization.

Apply principal component analysis to a breast cancer dataset, standardize features, derive two components, examine loadings, and visualize the separation between benign and malignant tumors in two dimensions.

Explore pre-processing techniques essential for research data, including data augmentation, SMOTE for imbalanced datasets, and principal components analysis to reduce dimensionality and improve model accuracy.

Explore supervised machine learning with convolutional neural networks, see how CNNs automatically extract image features, and learn their applications in image classification, object detection, and segmentation in research.

Learn how convolutional neural networks classify images by learning hierarchical features through convolutions and pooling. Prep datasets, train with loss function and optimizer, and tune hyperparameters for better results.

Train a convolutional neural network on the colorectal histology dataset of 5,000 images across eight classes; apply data augmentation, resize to 150 by 150, and evaluate with tf.keras.

Interpret CNN results by examining accuracy, overfitting, and a confusion matrix, and learn to use Grad-CAM to reveal image regions that drive adipose and tumor predictions.

Explore supervised machine learning with convolutional neural networks and their scientific research applications, how they work and why they are powerful, plus implementing a CNN in Python with interpreted results.

Explore unsupervised machine learning and the Louvain method for clustering data by building networks, maximizing modularity, and applying to gene expression research.

Apply the Louvain method to a Geo gene-expression dataset, build a k nearest neighbors network, and uncover two unsupervised clusters evaluated by silhouette score.

Compare unsupervised and supervised learning and explore real-world research applications like recommender systems and drug repurposing through the Louvain method evaluated with the silhouette score on NCBI gene expression data.

Apply machine learning to research by selecting high-quality data and suitable algorithms. Share findings through publications and presentations with clear visualizations.

Explore how machine learning applies to research, from data preprocessing with data augmentation, Smote, and PCA, to supervised and unsupervised algorithms like CNNs and the Louvain method.