Niclas Popp

ELLIS PhD Student at University of Tübingen and Bosch Center for Artificial Intelligence

[Niclas Popp @PoppNiclas]

Our findings on how knowledge distillation can be leveraged to effectively train image encoders on synthetic images have now been published in Transactions on Machine Learning Research (TMLR). The full text is now available at: https://t.co/y6crjo3elH

[Niclas Popp @PoppNiclas]

Big thanks to @jan_metzen and Matthias Hein for the great supervision and @Bosch_AI and the University of Tübingen for supporting this project.

[Niclas Popp @PoppNiclas]

Our findings on how knowledge distillation can be leveraged to effectively train image encoders on synthetic images have now been published in Transactions on Machine Learning Research (TMLR). The full text is now available at: https://t.co/y6crjo3elH

[Niclas Popp @PoppNiclas]

We show that feature distillation offers several advantages over contrastive vision-language training that jointly improve the transfer from synthetic to real images.

[Niclas Popp @PoppNiclas]

The work was carried out in the lab of @anto_scial together with Jonathan Fiorentino and Marco Stock

[Niclas Popp @PoppNiclas]

Happy to share STREAMLINE, a refined benchmarking strategy for GRN Inference Algorithms that focuses on the preservation of topological graph properties as well as the identification of hubs.

[bioRxiv @biorxivpreprint]

STREAMLINE: Structural and TopologicalPerformance Analysis of Algorithms for the Inference of Gene Regulatory Networks from Single-Cell Transcriptomic Data https://t.co/82jC9leqtW #bioRxiv

[Scialdone Lab @anto_scial]

How well can we estimate "structural properties" of gene regulatory networks (eg, robustness to perturbations, hubs) from single-cell RNA-seq data with the currently available algorithms?
We look at this in our new pre-print!
With @PoppNiclas @stk_mrc and Jonathan Fiorentino

[bioRxiv @biorxivpreprint]

STREAMLINE: Structural and TopologicalPerformance Analysis of Algorithms for the Inference of Gene Regulatory Networks from Single-Cell Transcriptomic Data https://t.co/82jC9leqtW #bioRxiv

[Niclas Popp @PoppNiclas]

Partitioning of an adjacency graph from a sparse matrix that occurs in thermoelastic models. The computations were done on Swedens (soon to be) largest super computer @KTHuniversity using a parallel implementation of Spectral Clustering: https://t.co/LLifDqYQch