Functional Triplet Motifs Underlie Accurate Predictions of Single-Trial Responses...
summary
High-speed two-photon targeted line scans sample up to ~350 excitatory layer 2/3 neurons at ~30 Hz and across datasets (~4,500 neurons) to predict single-trial calcium transients. A partial-correlation directed weight matrix (lags ±100 ms) plus linear summation of inputs uses nine 5-minute blocks (45 min) to predict a held-out 5-minute block, yielding ~80% mean variance explained (optimal regression ≈92%, inferred weights ~10% worse). Both tuned and untuned cells contribute, large weights and recurrent edges matter disproportionately — the simple network model works surprisingly well.
AI-written summary · about
chapters (15)
- 0:00Start
- 3:13GOALS
- 4:23OUTLINE
- 5:05OUR EXPERIMENTAL SET-UP
- 5:44RETINOTOPY TO CONFIRM VISUAL CORTEX
- 8:22CORTICAL V1 MICROCIRCUIT DYNAMICS IMAGED WITH HOPS
- 13:08MULTINEURONAL RESPONSES TO DRIFTING GRATINGS
- 15:22TUNED NEURONS SHOW VARIABLE SINGLE TRIAL ACTIVITY
- 16:30TUNING DOESN'T PREDICT SINGLE TRIAL RESPONSES VERY WELL
- 22:11BUILDING A FUNCTIONAL NETWORK USING PAIRWISE PARTIAL CORRELATION Edge weight - partial correlation w = 0.51
- 26:30FUNCTIONAL NETWORKS REFLECT TUNING IN THE POPULATION
- 27:01MODELING NEURON RESPONSES USING FUNCTIONAL NETWORKS
- 32:28POPULATION SIZE UNDERLIES PREDICTION ACCURACY
- 34:01RECURRENT CONNECTIONS ARE BIASED TOWARD LARGE EDGE WEIGHTS
- 35:28ASSEMBLIES
related talks
01h 13mComputation in Networks of Neurons in the Brain II
47mInterpreting Deep Neural Networks (DNNs)
58mStrong Generalization from Small Brains and No Training Data
13mEnhanced EEG Data Classification with 3D SNN and On-Line Learning ESN
27mWhat can artificial neural networks learn from biological neuromodulatory systems?
30mCracking the code: neuronal networks in the brain
45mMachine Learning, Deep Networks and Interpretability in Systems, Cognitive and...
43mThe Biological Path Towards Strong AI