Join our lab at https://t.co/5ffZIFT7sA, amazing PhD and Postdoc opportunities in NeuroImaging, Neurotech, ultrasound, computational neuroscience and neuromorphic computing, neuroimaging and much more!
Proud to have worked on this wonderful collaborative project with amazing colleagues! Brain WM affection in the cerebellum is associated with serum NfL in MS. #multiplesclerosis#MRI#disconnectome https://t.co/tzH2b9G4pJ
ArXiv https://t.co/8Yef0gBZB7: RL with sparse delayed reward: value-difference-based intrinsic reward. Outperforms both non-phasic RL & phasic SL baselines. Very similar to our RUDDER (Arjona-Medina et al) with redistributed reward, which uses differences of predicted Q-values.
Today is International Women in Engineering Day, which celebrates women in #engineering and campaigns for inclusivity within the sector.
Find out more at https://t.co/u2RdGORBUP #INWED22#WomeninSTEM
ArXiv https://t.co/hRrQd1Q6bI: Flat minimum search / sharpness aware minimization (SAM): measuring flatness by the largest eigenvalue λmax of Hessian. λmax is limited in explaining generalization. Confirms generalization benefits of large learning rates and SAM for minibatch SGD.
ArXiv https://t.co/Eg6izCPcyk: DOTA certified adversarial robustness using only off-the-shelf pretrained models. Denoised smoothing that uses a pretrained denoising diffusion model and a standard classifier. Improvements: 14% over certified SOTA, 30% over denoised smoothing.
ArXiv https://t.co/HLPPjUQn7y: Partial differential equations (PDEs): compares existing deep learning approaches and provide theoretical explanations for their performances. Novel methods that are more robust and lead to considerable performance improvements.
In this chapter, we review recent advances in data-driven and theory-informed computational models of networks for addressing marginalization in society and algorithms: https://t.co/Udxhry0EfE
Please RT and spread the word #OHBM2022, thank you!
The public vote for this years @OHBM#BrainArt Competition is now live - select your favorite for each category. 🧠🎨❤💚
👇🏻👇🏻👇🏻
https://t.co/NFOun3KFlD
Nahiyan's paper is out!
#Transfer#learning using knowledge from #Youtube videos could potentially aid in training deep learning models in the #data-#scarce brain imaging setting
https://t.co/Y1cmzbszRS
An eloquent profile of @nickyclayton22, whose pioneering corvid lab at @Cambridge_Uni is in danger of being shuttered. This would be a great loss & a major setback to the study of animal cognition and consciousness 👇🏽 https://t.co/5EljreQYYw