Mining Hubble Space Telescope images

We present a unsupervised machine learning technique to explore large sky surveys produced by imaging telescopes. Distinct from previous approaches this technique requires no pre-selection of target galaxy type; instead it automatically identifies objects that are similar. We apply the technique to the five Hubble Space Telescope CANDELS fields and compare the machine-based classifications to human-classifications from the Galaxy Zoo: CANDELS project. We find that although there is not a direct mapping between Galaxy Zoo and our hierarchical labelling, there is a good level of concordance between human and machine classifications. A catalogue and galaxy similarity search is provided for the CANDELS fields at http://www.galaxyml.uk/

Date: 24/03/2017
Time: 16:00
Location: LB252

Share this post on: Twitter| Facebook| Google+| Email

Categories

Tags

neural networks Associative memory Studentship cerebellum Artificial intelligence olfaction Dendritic computation Purkinje cell Neuronal Morphology Evolutionary algorithms Neuromorphic hardware Persistent homology Open position COVID-19 Network connectivity robotics Deep learning Convolutional Neural Networks Neurons computational neuroscience neuroscience computational modelling Homoeostasis Software development machine learning PCA Computer Science optimization Structural plasticity synaptic plasticity