Dark Machines

We are a research collective aiming to answer cutting edge questions about Dark Matter with the most advanced techniques that data science provides us with.

[Dark Machines @dark_machines]

Hello Twitter, this is the account of our brand new "Dark Machines" research collective. Follow us to know more about the application of #MachineLearning to #Physics and #Astronomy, and in particular to #darkmatter studies. More info at https://t.co/cWVZUtnAOK

[Dark Machines @dark_machines]

New @dark_machines related paper on automatic source identification.
"Training" on astronomical data from telescope A (Meerlicht) and application to telescope B (Hubble) works very well...
https://t.co/IFNE9uKUS9

[Sascha Caron @SaschaCaron]

Our @dark_machines LHC unsupervised data made it to an art project !!! This is really cool ;) Thanks.

with @CatDogLUofM @xmpierinix , @BryanOstdiek , @l_hendriks , @MCvBeekveld and many others

[Sascha Caron @SaschaCaron]

This is the link to the nice CERN-Courier article about @dark_machines and LHC Olympics anomaly detection challenges at LHC.
https://t.co/ofvziwJlqR

with e.g. @BPNachman , @MCvBeekveld , @BryanOstdiek , @l_hendriks , @xmpierinix , @CatDogLund, @GregorKasieczka

[Dark Machines @dark_machines]

You are looking for training data for your machine learning model.
Take the simulated LHC events, full SM background + signal challenge.
> 1000000000 events
https://t.co/Y3zAlv01DO

[Dark Machines @dark_machines]

This project proposes adding such algorithms to LHC searches (and beyond!) to define signal regions based on anomaly score to detect unexpected new physics.

[Dark Machines @dark_machines]

Today on arxiv:

The Dark Machines Anomaly Score Challenge:

Benchmark Data and Model Independent Event Classification for the Large Hadron Collider

https://t.co/j7MyxDLH9T

[Dark Machines @dark_machines]

Interesting is that the winners all have a fixed target loss function in one way or the other (and OR this with the likelihood of a flow model).

[Dark Machines @dark_machines]

This work also includes a "data challenge" (like all @dark_machines projects).

All data available on @ZENODO_ORG (see links in paper).

Can you help to (automate) the identification of (astronomical) sources ?

[Dark Machines @dark_machines]

"Automatic" identification/catalog for point sources project

started in @dark_machines made it to arxiv today:
https://t.co/bvduJPG46v

Analysis chain:

gamma ray images
--> u-nets+kmeans or centroid-nets
--> Bayesian classification
--> astro catalogue with posteriors

[Dark Machines @dark_machines]

@GregorKasieczka Would have been great to collaborate.

[Dark Machines @dark_machines]

Secret dataset for the unsupervised search challenge available ... stay tuned...

[Dr. Melissa van Beekveld @MCvBeekveld]

@SaschaCaron You can find it here: https://t.co/7M6IHpcsiT 🤗

[Dark Machines @dark_machines]

@GregorKasieczka Two timescales:

a) 1-2 weeks for https://t.co/S7DJCqywwF challenge

b) Indefinitely you can try to find the anomalous events, submit your anomaly score (0 to 1 per event) to the "Secret Dataset Team, Melissa, Zhonghy et al." and use the evaluation (e.g. AUC) for your work.

[Bob Stienen @bob_stienen]

A new paper appeared! The @dark_machines sampling group has put its latest work on @arxiv: A comparison of optimisation algorithms for high-dimensional particle and astrophysics applications.

Paper can be found at https://t.co/NuuIyTHBkP. Summary thread below 👇

[Sascha Caron @SaschaCaron]

First in a series of 2021 @dark_machines papers:

A comparison of optimisation algorithms for high-dimensional particle/astrophysics applications

Trying to find optimal parameters of your physics model ?
Is your chain of simulators non-differentiable?

->https://t.co/z16bAGymm6

[Gianfranco Bertone @gfbertone]

Interesting @EmoryTMLS virtual workshop on “Learning New Physics with Machine Learning” starting today at 4pm CEST 👇

[Dark Machines @dark_machines]

We have decided in a meeting of the organisers to postpone the 2020 darkmachines workshop at CERN until autumn (dates still have to be set) due to the current situation with SARS-COV-2.

Please cancel your CERN hostel on an individual basis.
Sorry !

[Sascha Caron @SaschaCaron]

LES HOUCHES 2019: PHYSICS AT TEV COLLIDERS
NEW PHYSICS WORKING GROUP
is now online

https://t.co/FWx13ArGLB

Comes with 10fb-1 of simulated LHC data for phenomenology from @dark_machines
+ other pheno ML training data

First datasets https://t.co/z5M95kfsUY

[EuCAPT @EuCAPT]

We invite all scientists (PhD students, postdocs, and staff) active in theoretical astroparticle physics and cosmology & affiliated to a research institution in a European country, to participate in the 1st EuCAPT census by following this link:
https://t.co/CU00rAvDB9 Please RT