Spatial data upscaling

Pipeline for augmenting sparse data for genetic optimisation

Augmenting data in specific search space using large broad dataset

During data collection for genetic optimisation, available data may be broad instead of specific. Training a neural network on broad data can enable data collection for a specific search space at increased resolution. This pipeline suggests a method for training a model on a broad dataset, and applying it to generate data at a desired resolution for a specific search space.

Broad dataset has large amount of data..	..but is insufficient at specific search space.	The model wants to learn spatial features of the datset..	..and is able to predict similar data, suggesting it has learnt the dataset.	Using this model, we can generate high resolution data for our search space.

Blog post

Read about this project in further detail at https://cutwell.github.io/neural-network-augmentation/.

Pipeline

1. Follow train_model.ipynb to train a regression model.
2. View run_model.ipynb for ways to apply the regression model to predict wind data for the search space.