CNBP

Pascal Vincent start off the day 2 of #main2018 with a good gentle introduction of deep learning and good practical tips.

#main2018 @crocodoyle it seems like you got some serious competitions in terms of the quality control aspect. :) poster 25. How come you two are not collaborating?

#MAIN2018 speakers of Day 1. Pretty diverse talks than expected. Looking forward to see Day 2

Memes are more efficient than genes. Interesting bits from Dr. Bengio's #MAIN2018 talk on ANN and its biological origins. i.redd.it/xmulu1f6urt11.…

Hardcore.... to study how rodent process whisker sensory signals @dyamins at #main2018 mentioned building a biomechanical model of rodent whiskers...and then CNN models of processing the biomechanical signal. Please do share pictures.

@leavittron @MAIN_Community The work of @dyamins presented at #MAIN2018 has very interesting implications tieing down the interpretation of the CNN and its biological parallel in visual system. It has interesting relevance from both CNN #interpretability and CNN network architectures designs. Very inspiring

Dan Yamins tearing through this talk at 800wpm is exactly what I need when the tendrils of lunch are trying to drag me into sleep. His speed to comprehensibility ratio is an insane ~ 1:1 (factored from 800:800) #MAIN2018 @MAIN_Community

Also thay is some slick slick manifold visualization.

Super cool visual (ha) parallel drawn by @dyamins of convolutional neural networks and visual system at #MAIN2018.

Thank @Mehravehs for her concise, well explained talk on state specific brain parcellation at #MAIN2018 .

Some interesting quotes collected by Dr.Yves Fregnac at #MAIN2018 especially the quote on the key differences between insight and data.Well crafted experiments and careful interpretation are still extremely crucial to help us understand neuroscience and any scientific experiment

Such a big team. Looking forward to learn more about Neuromod. What is the website? # twitter handle? @pierre_bellec

#MAIN2018 cool announcement by @pierre_bellec of neuromod. No hashtag and twitter account yet? Hooray for #openscience and #reproducibility. Looking forward to the high quality data release to test the robustness of our data analyses pipelines.

Interesring point from @pkdouglas16 on the importance of noise in biological and artificial neural networks. Gotta do more dropout 🤣.... #MAIN2018 . Thanks for the rich and indepth talk...

Excellent summary slides from @danilobzdok at #MAIN2018 Thanks so much! He also touched on the important tradeoff of interpretability versus prediction accuracy of modeling. Clinical applications versus research understanding have quite different statistical needs.

Interesting point at #MAIN2018 about how predictive significance might not be group significant. Always assumed that is a given. still have much to read.... also, super interesting timeline on the history of statistical methods. Of note, GLM and 10fold CV are only since 1975.

@danilobzdok mentiomed in #MAIN2018 important differences between results that are statistically significant versus results that enable case level prediction. Quite an important distinction especially in many production level integration of ML. P << 0.05 is seldom enough.

Kudos to Dr. Evans from #hbhl for giving a cool level headed perspective of the ML/DL application in neuroscience and the long road ahead of us. Looking forward to see some concrete useful application of ML/CV/RL/DL that we can adopt from the works presented at #MAIN2018

@crocodoyle Looking forward to your talk. Don't let us down. We hyped about your talk among us for a while...

@angelatamtweets's #MAIN2018 tweet all the way from Singapore got a shoutout at the opening ceremony by the organizer. 🤣. Surprised singapore has no epic bagels beside all the delicious food as experienced during #ohbm2018

We are at #MAIN2018! I missed the desk and power charger from #main2017 though. Bright and early Sunday with tons of coffee...looking forward to see how people put computer vision to work! Wonder how much reinforcement learning there will be versus CNN. Maybe capsule too?