RT @bttyeo: 1/11 Excited to share our @Naturestudy led by @Leon_Oo1 @csabaorban @ZShaoshi. It is well-known that A….

RT @TrendsCognSci: The interoceptive origin of reinforcement learning. Review by Lilian Weber (@lilwebian), Debbie Yee (@debyeeneuro), Dana….

RT @marcelomattar: Thrilled to see our TinyRNN paper in @nature! We show how tiny RNNs predict choices of individual subjects accurately wh….

Single figure for the Five Breakthroughs of human intelligence (

The book <A brief history of intelligence> is incredible! I really enjoyed reading, and it inspires me a lot about many research topics. This book describes how natural and artificial intelligences have developed throughout the history. Highly recommend! (right: Korean version)

RT @BangL93: 🧠264 pages and 1416 references chart the future of Foundation Agents. Our latest survey dives deep into agents—covering brain….

RT @gershbrain: New survey on predictive representations in reinforcement learning, covering AI, cognitive, and neuroscience perspectives:….

RT @DrBMunn: I am absolutely stoked to share our new study on multiscale neural dynamics across species and behaviour + network simulation….

RT @PAINthejournal: Now in #PAIN: “Decoding pain: uncovering the factors that affect the performance of neuroimaging-based pain models” by….

Starting to work in Nagai lab from this week! Luckily, I had an opportunity to experience the demonstration for a robot and VR because of the open lab for high school students on my day 2. Exciting💥

RT @choongwanwoo: I'm pleased to share our lab’s collective effort: ‘Pain Neuroimager Manifesto: Towards Person-Centered Neuroscience of Pa….

Shoutout to my advisor @choongwanwoo and our center @CNIR_IBS for the great support! 🌟 (15/15).

These findings will serve as a useful reference for making decisions on neuroimaging-based biomarker development, highlighting the importance of a careful selection of modeling variables to build better-performing neuroimaging pain biomarkers. (14/15).

The final benchmark analysis focused on the impact of sample size. While increasing the sample size, we developed new models and calculated performance measures. The results showed that increasing sample sizes improved model performance, consistent with previous studies. (13/15)

The third benchmark compared idiographic vs population-level predictive modeling. Results showed no advantage for idiographic models, contradicting the belief that personalized models perform better by reducing between-individual variability. (12/15)

Our results showed that including more brain regions improved model performance, consistent with previous studies. However, including extra brain regions beyond the regions known to be important for pain prediction did not always guarantee higher model performance. (11/15).

The second benchmark was on spatial scale. We repeated the model development 100 times with random brain region combinations. For brain-wide analysis, we used three masks: 21 pain-predictive regions, Neurosynth “Pain,” and Gray matter. (10/15)

The first benchmark analysis focused on data levels, indicating the number of trials averaged for model training and testing. Less averaging improved model performance for training data while more averaging for testing data did not. (9/15)

Different experimental designs in studies hindered direct comparisons. We collected and analyzed a large-scale local pain fMRI dataset using a single design, focusing on four key aspects. We aimed at binary classification of high vs low pain and prediction of pain ratings. (8/15).

We compared model performances across different targets and options. We focused on six aspects: prediction tasks, modeling targets, data and model levels, spatial scale, and sample size. These key aspects guide future studies and are included in the benchmark analysis. (7/15)