Most ML interpretation methods have a common enemy: Correlated features. They ruin interpretation both on a technical and a philosophical level. Why correlation is problematic, how to patch it, and why we have no cure. A thread 🧵

If you are ECMTB2022 #smb2022 and are interested in #mathonco don’t miss presentations from our lab Oslo Centre for Biostatistics and Epidemiology: Even Moa Myklebust @[email protected], today 17:50 at DKFZ big Tuğba Akman, Thursday 10:20 at DKFZ big Also you should have a look at the poster by Youness Azimzade

Closing this 1st contributed session is Even Moa Myklebust @[email protected] from Alvaro Köhn-Luque lab discussing, Phenotypic deconvolution in heterogeneous cancer cell populations using drug screening data #SMB2022 ECMTB2022

Explained in class today that delta means a small positive number you choose and epsilon means a small positive number your enemy chooses.

Massive thanks to #JasmineFoo and #KevinLeder for hosting us University of Minnesota It has been a wonderful and productive visit. @freeges Even Moa Myklebust @[email protected] Oslo Centre for Biostatistics and Epidemiology

Ran one of our essay questions through OpenAI's new chatbot. Essays are dead. Back to hand-written exams I guess. Sigh.

Finally!! Very happy to see this that the paper we have been working on for so long (two and a half years in my case) is finally published! Go check it out (especially if you're interested in drug response heterogeneity in cancer).

We have a new #mathonco preprint led by #ChenyuWu & #KevinLeder from University of Minnesota Using birth-death processes to infer tumor subpopulation structure from live-cell imaging drug screening data arxiv.org/abs/2303.08245 Even Moa Myklebust @[email protected] @freeges Oslo Centre for Biostatistics and Epidemiology Dagim jorrit enserink

Very happy that my joint work with #HelleSkjerven and Vestre Viken is out! We compared simple and skin-sparing mastectomy of ductal carcinoma in situ in a retrospective register-based cohort study.

Panel discussion on modeling tools and the future of digital medicine Hanna Hübner Youness Azimzade Even Moa Myklebust @[email protected] @ltronneberg

Had such a great time at the RESCUER meeting in Venice! Grateful for the people, the experience and the science!!

Great work by Youness, happy to see this out!!

Our guys from the cancer modeling team Oslo Centre for Biostatistics and Epidemiology at the Nordic Precision Cancer Medicine in Holmenkollen Youness Azimzade Even Moa Myklebust @[email protected] L

Have you ever done a dense grid search over neural network hyperparameters? Like a *really dense* grid search? It looks like this (!!). Blueish colors correspond to hyperparameters for which training converges, redish colors to hyperparameters for which training diverges.

To every political leader and journalist making and writing speeches: You can't save Navalny. You *can* still save Assange. If you're silent here, when and where you can actually influence the outcome, you were never standing up for principles—you were just hoping for applause.

Why can't academics give up flying? We identified cultural, individual, and organizational reasons academics think we can't decarbonize research production. Especially relevant as more conferences & workshops return to offline-only formats. nicholaspoggioli.files.wordpress.com/2023/08/poggio…

Just out! How well can relapse in multiple myeloma be predicted using previous M protein measurements and baseline covariates? We develop statistical methods and test them on 229 patients from the IKEMA trial. #mathonco

New preprint: In a simulated study, we predict treatment response and relapse in cancer with a hierarchical Bayesian model of subpopulation dynamics: arxiv.org/abs/2405.14508 Alvaro Köhn-Luque @freeges Oslo Centre for Biostatistics and Epidemiology