1/5 Our new preprint https://t.co/2WYKy1ZEoi uses meta-learning to boost functional connectivity based prediction of *NEW* phenotypes in small samples. We achieve 100%-400% improvements in many scenarios!
@hetong93 @AvramHolmes @anlijuncn@danilobzdok @AvramHolmes @INM7_ISN
2/2 all algorithms dramatically improved when sample size increased from 100 to 1000. Improvement was smaller, but still significant, when sample size increased from 1000 to 5000 subjects. KRR was competitive at all sample sizes.
Publicly available code: https://t.co/1tf5x0rilU
1/2 Updated preprint (https://t.co/zvTao5Up3J) showing that kernel regression and DNNs have equal performance for RSFC behavioral prediction. In this version, we found that ...
@hetong93@rubykong92@danilobzdok@INM7_ISN@mertrory @AvramHolmes
Preprint: https://t.co/kZrpYlgPLF We compare kernel regression with deep neural networks (DNNs) for functional connectivity behavioral prediction using ~10,000 participants. Hint: DNNs do not perform better
@hetong93@rubykong92@danilobzdok@INM7_ISN@mertrory @AvramHolmes
2/4 #OHBM2018
#1090: @SunNanbo found latent AD factors underlying cognitive, atrophy, tau heterogeneity
#1172: @tangsy93 found latent Autism factors with dissociable RSFC & behavior
#1411: @valeria_kebets found latent behavioral and RSFC components across mental disorders
4/4 #OHBM2018
#2259: Minh shows that recurrent neural networks can predict AD progression 5 years into the future
#2499: @Jadecci shows off her MNI152-fsaverage mapping from this thread https://t.co/Mum6MtN1AV
1/4 Come by our #OHBM2018 posters & discuss!
#2380: @rubykong92 presents her individual-specific network parcellations from this thread: https://t.co/2mlRTNqnk8
#2530: Dragana presents a new multisubject stochastic blockmodels with subject-varying cluster assignment
3/4 #OHBM2018
#2621 Rajan shows that individual-specific resting-fMRI subspaces improve behavioral prediction
#2391 @hetong93 compares deep learning & kernel regression. PRNI best paper runner-up!
#2370 @RLiegeois_Neuro shows that dynamic & static FC encode distinct behaviors