These references are to the only work with which I am intimately familiar: my own. I will be happy to also post references to other works if both the reference and the description are provided by a named source, but I have neither the time, the interest, nor the ability to both understand each work anyone suggests and to defend my interpretation of it. Sorry.
Paper | Year | Main point | Weakness | Similar work |
---|---|---|---|---|
95 Hypotheses on the Informational Structure of Life which are Multi-scale, Human-readable, Provacative, Plausible, Coherent, Actionable, and Falsifiable | 2017 | Proposes nine generic, testable information-processing mechanisms, ranging from self-replication to self-calbration, by which Life functions and grows. | This theory ought to be self-evident. No equations, data, references, or experimental detail. | |
Elastic Nanocomputation in an Ideal Brain | 2014 | Explains continuous low-dimensional representations, and calculates the millions-fold advantages of continuous simulatrix over "neural" representations. | First draft, few refs | |
Brain-like software architecture | 2002 (unpublished) | Modular architecture of compressor/predictors | No 3-D priors | Numenta |
Unsupervised Pixel-prediction | 1996 (NIPS) | Hebbian system learns 2-D motion from structured input, and learns further features from prediction failures. | Must learn the dimensionality of its compressed representation (i.e. 2) | Rao & Ballard, Hawkins |
Modeling thalamus as a predictive comparator | 1993 (unpublished) | Models thalamus as an I/O port subtracting predictions from new input. | about single-neuron dynamics, not overall function | |
Simple Codes vs. Efficient Codes | 1993 | Contrasts the traditional assumption that neurons use slow and noisy rate-codes with the alternative that they use high-precision single-spike codes, and shows how the single-spike code is in fact consistent with experiment. | Only addresses info density, not function | Shadlen and Newsome |