Target group
- Data Scientists in research and industry
Your requirements
You have data and first need unbiased access to them. You ask yourself: “Do I have enough data?”, “What is the data quality?”, “Will the data help me with my task?”Our offer
Unsupervised learning methods can be used to estimate the complexity of the data and, if possible, to represent them in a compact fashion, e.g. via suitable representatives. Cluster methods or approaches to dimension reduction are suitable here.
We offer both simple and complex procedures to build an understanding of your own data. This includes classic methods such as k-means clustering, mean shift, a PCA or simple autoencoders.
Depending on the question, even sparsely populated solutions can be optimized or meta information can be incorporated to structure ML learned representations. We also offer solutions for anomaly detection, e.g. based on normalizing flows.
Requirements
Basic knowledge in statistics and programmingSuccess stories
Various components have already been successfully published:- https://www.tnt.uni-hannover.de/papers/data/1553/CMS.pdf
- https://arxiv.org/abs/2008.12577
- https://arxiv.org/abs/2110.02855
- https://github.com/creinders/ClusteringAlgorithmsFromScratch
- https://github.com/marco-rudolph/differnet
- https://github.com/marco-rudolph/cs-flow
