Allen Institute for Neural Dynamics reposted this
🚀 Exciting Opportunity Alert! 🚀 Are you passionate about #Neuroscience and #PredictiveCoding? 🧠🔍 We invite all neuroscientists to contribute to an open science experiment co-organized by the OpenScope program Allen Institute, Michael Berry, Colleen Gillon, Konrad Kording, and the CCN workshop! 🤝🌐 We're exploring predictive coding and need YOUR expertise to review the computational and experimental literature. 📚 💡 Predictive coding is a leading theory of cortical function, with significant implications for perception, cognition, learning, and overall brain function. 🧠✨ ❓ Central question: What mechanisms underlie the subtraction of predictions from sensory input data? 📄 We've created a Google Doc for collaborative brainstorming on a future perspective piece. Dive in, share your insights, and help shape the future of neuroscience! 📝🌟 https://lnkd.in/gXSQmbFz 🔗 The Allen Institute’s OpenScope Program is READY to perform the proposed experiment on the @alleninstitute pipelines. 🌐🚀 🌍 This is #ScienceInTheOpen! Your contributions will be credited, and you’ll be part of an innovative community pushing the boundaries of knowledge. 🌟📈 We will meet in Boston on August 8th as part of the CCN workshop to discuss new experiments that the field needs to move forward on Predictive Coding. Attend this event. https://lnkd.in/gJAzV7hF 📊 Datasets from this endeavor will be made immediately openly accessible to the entire scientific community for analysis. 📊🌐 💬 Questions? DM us or comment below! Let's actually DO science in the Open 🔍 #OpenScience #CollaborativeResearch #CCN #Neuroscience
The Free Energy Principle & Deep Learning | The Fusion of Variational Methods and AI to: 1) Demystify Human Perception 2) Demystify Neuro-Psychiatric Disorders 3) Create True AI
1moThe main problem with the predictive coding (PC) literature (at least with most of it) is that the continuous evidence accumulation (using Generalised filtering) of PC has been totally ignored and the main point of focus has been local Hebbian learning (as opposed to Backpropagation in #deeplearning). Furthermore, a proper uncertainty estimation over the states and observation seems to have been ignored for the most part. This is the result of my personal review of the PC literature. The subset of the literature that has done PC justice, is what you typically find done by Rick Adams and Karl Friston as main authors. Interestingly I can see only 2 references to Friston and 1 reference to Adams in the document that you have shared. Anyway, I hope this is helpful :-)