The Earth-Life Science Institute at the Tokyo Institute of Technology held a Machine Learning and Deep Learning Bootcamp in collaboration with Machine Learning Tokyo for scientists in Biology, Chemistry, Astro- and Geophysics, and other fields researching the origins of life. Find the MLT materials on GitHub:

 

Slides

ML Research Project Management

Intro to Deep Learning

Intro to Convolutional Neural Networks

 

Notebooks

Exploratory Data Analysis

Data Visualization

Train a Convolutional Neural Network

 

Data: Kaggle – DeepSat (SAT-6) Airborne Dataset
405,000 image patches each of size 28×28 and covering 6 landcover classes

• Each sample image is 28×28 pixels with 4 bands – red, green, blue, near infrared.
• The training and test labels are one-hot encoded 1×6 vectors
• The six classes represent the six broad land covers which include barren land, trees, grassland, roads, buildings and water bodies.
• Training and test datasets belong to disjoint set of image tiles.
• Each image patch is size normalized to 28×28 pixels.
• Once generated, both the training and testing datasets were randomized using a pseudo-random number generator.