Unifying the Theories of Neural Information Encoding

New theory gives concrete predictions for previously unstudied coding regimes.


Digital Cameras have the capacity to record in extraordinary detail, however sparing every one of that information would take up an immense space. Thus, it leads to an issue, how to compress a video—that is, expel data—such that we can’t see the distinction when it is played back?

Likewise, as we approach our daily lives, our eyes are overwhelmed with visual data, however, the neurons in our eyes have certain requirements—simply like information engineers. Along these lines, given this rich arrangement of boosts, how do neurons choose what to remove and send on to the mind? Neuroscientists have been making this inquiry for a considerable length of time and had beforehand utilized a few unique speculations to clarify and foresee what neurons will do in specific circumstances.

Now, a scientist, scientists at the IST Austria, have developed a framework that unites the previous theories as special cases, and enables them to make predictions about types of neurons not previously described by any theory.

One of the primary objectives of sensory neuroscience is to anticipate neural reactions utilizing scientific models. Already, these forecasts depended on three fundamental hypotheses, each of which had an alternate territory of pertinence, relating to shifting suspicions about the neurons‘ inside limitations, the sort of flag, and the motivation behind the accumulated data.

Generally, a neural code predicts when a neuron should “fire”, that is, emit an action potential signal much like a digital “1” in the binary alphabet that our computers use. Such collection of one or more neurons firing at particular times can thus encode information.

Productive coding expects that the neurons encode however much data as could be expected, given their inward limitations (clamor, digestion, and so forth). Predictive coding, then again, expect that lone the data pertinent to foreseeing the future (e.g. which way a creepy crawly will fly) is encoded. At last, Sparse coding expect that exclusive a couple of neurons are dynamic at any one time.

Gašper Tkačik, IST Austria Professor said, “One problem with this situation was that it was unclear how these theories were related, and if they were even consistent with each other. These latest developments bring order to the theoretical landscape. Before, there was no clear notion of how to connect or compare these theories. Our framework overcomes this by fitting them together within an overarching structure.”

In this newly developed framework, a neural code can be deciphered as the code that expands a specific scientific capacity. This capacity and therefore, the neural code augmenting it—relies upon three parameters: the clamor in the flag, the objective or task and the complexity of the signal being encoded.

Tkačik explained, “The theories described above are valid only for specific ranges of values for these parameters and do not cover the entire possible parameter space, which presents problems when trying to test them experimentally. When you design stimuli to present to the neurons to test your model, it is extremely difficult to distinguish between a neuron that is not fully consistent with your favorite theory or the alternative where your favorite theory is simply incomplete. Our unified framework can now give concrete predictions for parameter values that fall in between the previously studied cases.”

The group’s brought together hypothesis conquers before constraints by enabling the neurons to have “blended” coding targets; they don’t need to fall into one clear, beforehand contemplated class. For instance, the new hypothesis can cover the situation where neurons are separately exceptionally uproarious yet at the same time ought to productively encode scanty jolts.

All the more by and large, ideal neural codes can be set on a continuum agreeing the parameter esteems that characterize limitations to optimality, which clarifies marvels that were already watched, yet not clarified by any of the current models.

First author Matthew Chalk said, “A lot of the theories that give predictions tend to be inflexible when tested: either it predicted the correct outcome or it didn’t. What we need more of, and what our paper provides, are frameworks that can generate hypotheses for a variety of situations and assumptions.”

Notwithstanding enriching the hypothesis with more prominent adaptability, their structure gives solid expectations for kinds of neural encoding that were already unexplored, for instance, encoding that is both inadequate and prescient. To catch up on the hypothesis created in their paper, Chalk is outlining investigations to test these forecasts and help sort neurons as proficient, prescient, or scanty—or as a mix of these coding targets. In Olivier Marre’s lab at the Institut de la Vision in Paris, he concentrates on the retina, and is creating visual boosts that will actuate retinal neurons in order to best uncover their coding destinations.

Tkačik said, “And the framework can also be applied more broadly: You don’t necessarily need to think about neurons. The idea of framing this problem in terms of optimization can be used in any sort of signal processing system, and the approximation allows us to study systems that would normally have computationally intractable functions.”

Matthew Chalk, Olivier Marre, and Gašper Tkačik: “Towards a unified theory of efficient, predictive and sparse coding”, PNAS 2017