Bassett 2018

Bassett, D. S., Zurn, P., & Gold, J. I. (2018). On the nature and use of models in network neuroscience. Nature Reviews Neuroscience, 1.

These days "a model" can mean many different things. Great figure with a summary of typical graph questions (communities, hubs, etc.) and extensions (dynamics, simplices). Hierarchy of models: descriptive (when graph makes it easy); predictive (dynamics, or edge dynamics). Then introduces another 3D classification: data vs theory, coarse vs low-level, and functional vs structural.

Greenland 2016

Greenland, S., Senn, S. J., Rothman, K. J., Carlin, J. B., Poole, C., Goodman, S. N., & Altman, D. G. (2016). Statistical tests, P values, confidence intervals, and power: a guide to misinterpretations. European journal of epidemiology, 31(4), 337-350.

Great summary of neo-Fisherian interpretation of p-values (amont other things).

Gilpin 2018

Gilpin, W. (2018). Cellular automata as convolutional neural networks. arXiv preprint arXiv:1809.02942.

Tried to predict cellular automata with deep networks (ReLU). Curious measure of entropy of activation patterns for almost-binary networks (just go through all possible inputs, look at the distribution of activation patterns, calculate H). But then at later layers entropy drops, and it drops faster if automata rules are simpler.

Thompson 2016

Thompson, A. W., Vanwalleghem, G. C., Heap, L. A., & Scott, E. K. (2016). Functional profiles of visual-, auditory-, and water flow-responsive neurons in the zebrafish tectum. Current Biology, 26(6), 743-754.

Ca imaging with selective plane microscopy. Good refs on cell ensembles. They find ensembles that selectively encode stimulus features. PCA+promax, then clustering to identify them. Classification depends on threshold and manual elimination of "artifactual clusters", so not too objective. Inputs for each cluster tend to be in same laminae (layers). For moving bars, not much correlation with position, suggesting that it's not retinotopy, but true directionality. Most cells are unimodal (only 1-3% belong to two clusters).

Implications: Consider that there are many cells with different sensory / multisensory properties, so it is only to be expected that they would be tuned differently, and possibly via different mechanisms. Especially if they are starting to segregate into layers, as this would imply different distances from the soma to the dendrites and (potentially) the axon initial segment.

Golowasch 2002

Golowasch J, Goldman MS, Abbott LF, Marder E (2002) Failure of averaging in the construction of a conductance-based neuron model. J Neurophysiol 87:1129–1131 

If one averages cell parameters, and plugs them in the model, the resultant "average cell" spikes entirely differently than original cells, as parameters are no longer coordinated. But does it mean by extension that comparing averages of underlying parameters across groups may not be the best way to detect effects of experimental manipulations?

Prinz 2004

Prinz, A. A., Abbott, L. F., & Marder, E. (2004). The dynamic clamp comes of age. Trends in neurosciences, 27(4), 218-224.

A good one-paragraph summary of how injecting current in the soma is different from currents generated in the dendrites, and that it's a problem. No refs though.