Neural coding

From Wikipedia, the free encyclopedia

 

Neural coding is a neuroscience field concerned with characterising the hypothetical relationship between the stimulus and the individual or ensemble neuronal responses and the relationship among the electrical activity of the neurons in the ensemble. Based on the theory that sensory and other information is represented in the brain by networks of neurons, it is thought that neurons can encode both digital and analog information.

Overview

Neurons are remarkable among the cells of the body in their ability to propagate signals rapidly over large distances. They do this by generating characteristic electrical pulses called action potentials: voltage spikes that can travel down nerve fibers. Sensory neurons change their activities by firing sequences of action potentials in various temporal patterns, with the presence of external sensory stimuli, such as light, sound, taste, smell and touch. It is known that information about the stimulus is encoded in this pattern of action potentials and transmitted into and around the brain.

Although action potentials can vary somewhat in duration, amplitude and shape, they are typically treated as identical stereotyped events in neural coding studies. If the brief duration of an action potential (about 1ms) is ignored, an action potential sequence, or spike train, can be characterized simply by a series of all-or-none point events in time. The lengths of interspike intervals (ISIs) between two successive spikes in a spike train often vary, apparently randomly. The study of neural coding involves measuring and characterizing how stimulus attributes, such as light or sound intensity, or motor actions, such as the direction of an arm movement, are represented by neuron action potentials or spikes. In order to describe and analyze neuronal firing, statistical methods and methods of probability theory and stochastic point processes have been widely applied.

With the development of large-scale neural recording and decoding technologies, researchers have begun to crack the neural code and have already provided the first glimpse into the real-time neural code as memory is formed and recalled in the hippocampus, a brain region known to be central for memory formation. Neuroscientists have initiated several large-scale brain decoding projects.

Encoding and decoding

The link between stimulus and response can be studied from two opposite points of view. Neural encoding refers to the map from stimulus to response. The main focus is to understand how neurons respond to a wide variety of stimuli, and to construct models that attempt to predict responses to other stimuli. Neural decoding refers to the reverse map, from response to stimulus, and the challenge is to reconstruct a stimulus, or certain aspects of that stimulus, from the spike sequences it evokes.

Hypothesized coding schemes

A sequence, or 'train', of spikes may contain information based on different coding schemes. In motor neurons, for example, the strength at which an innervated muscle is contracted depends solely on the 'firing rate', the average number of spikes per unit time (a 'rate code'). At the other end, a complex 'temporal code' is based on the precise timing of single spikes. They may be locked to an external stimulus such as in the visual and auditory system or be generated intrinsically by the neural circuitry.

Whether neurons use rate coding or temporal coding is a topic of intense debate within the neuroscience community, even though there is no clear definition of what these terms mean. In one theory, termed "neuroelectrodynamics", the following coding schemes are all considered to be epiphenomena, replaced instead by molecular changes reflecting the spatial distribution of electric fields within neurons as a result of the broad electromagnetic spectrum of action potentials, and manifested in information as spike directivity.

Rate coding

The rate coding model of neuronal firing communication states that as the intensity of a stimulus increases, the frequency or rate of action potentials, or "spike firing", increases. Rate coding is sometimes called frequency coding.

Rate coding is a traditional coding scheme, assuming that most, if not all, information about the stimulus is contained in the firing rate of the neuron. Because the sequence of action potentials generated by a given stimulus varies from trial to trial, neuronal responses are typically treated statistically or probabilistically. They may be characterized by firing rates, rather than as specific spike sequences. In most sensory systems, the firing rate increases, generally non-linearly, with increasing stimulus intensity. Any information possibly encoded in the temporal structure of the spike train is ignored. Consequently, rate coding is inefficient but highly robust with respect to the ISI 'noise'.

During rate coding, precisely calculating firing rate is very important. In fact, the term "firing rate" has a few different definitions, which refer to different averaging procedures, such as an average over time or an average over several repetitions of experiment.

In rate coding, learning is based on activity-dependent synaptic weight modifications.
Rate coding was originally shown by ED Adrian and Y Zotterman in 1926. In this simple experiment different weights were hung from a muscle. As the weight of the stimulus increased, the number of spikes recorded from sensory nerves innervating the muscle also increased. From these original experiments, Adrian and Zotterman concluded that action potentials were unitary events, and that the frequency of events, and not individual event magnitude, was the basis for most inter-neuronal communication.

In the following decades, measurement of firing rates became a standard tool for describing the properties of all types of sensory or cortical neurons, partly due to the relative ease of measuring rates experimentally. However, this approach neglects all the information possibly contained in the exact timing of the spikes. During recent years, more and more experimental evidence has suggested that a straightforward firing rate concept based on temporal averaging may be too simplistic to describe brain activity.

Spike-count rate

The spike-count rate, also referred to as temporal average, is obtained by counting the number of spikes that appear during a trial and dividing by the duration of trial. The length T of the time window is set by the experimenter and depends on the type of neuron recorded from and to the stimulus. In practice, to get sensible averages, several spikes should occur within the time window. Typical values are T = 100 ms or T = 500 ms, but the duration may also be longer or shorter.

The spike-count rate can be determined from a single trial, but at the expense of losing all temporal resolution about variations in neural response during the course of the trial. Temporal averaging can work well in cases where the stimulus is constant or slowly varying and does not require a fast reaction of the organism — and this is the situation usually encountered in experimental protocols. Real-world input, however, is hardly stationary, but often changing on a fast time scale. For example, even when viewing a static image, humans perform saccades, rapid changes of the direction of gaze. The image projected onto the retinal photoreceptors changes therefore every few hundred milliseconds.

Despite its shortcomings, the concept of a spike-count rate code is widely used not only in experiments, but also in models of neural networks. It has led to the idea that a neuron transforms information about a single input variable (the stimulus strength) into a single continuous output variable (the firing rate).

There is a growing body of evidence that in Purkinje neurons, at least, information is not simply encoded in firing but also in the timing and duration of non-firing, quiescent periods.

Time-dependent firing rate

The time-dependent firing rate is defined as the average number of spikes (averaged over trials) appearing during a short interval between times t and t+Δt, divided by the duration of the interval. It works for stationary as well as for time-dependent stimuli. To experimentally measure the time-dependent firing rate, the experimenter records from a neuron while stimulating with some input sequence. The same stimulation sequence is repeated several times and the neuronal response is reported in a Peri-Stimulus-Time Histogram (PSTH). The time t is measured with respect to the start of the stimulation sequence. The Δt must be large enough (typically in the range of one or a few milliseconds) so there are sufficient number of spikes within the interval to obtain a reliable estimate of the average. The number of occurrences of spikes n_K(t;t+Δt) summed over all repetitions of the experiment divided by the number K of repetitions is a measure of the typical activity of the neuron between time t and t+Δt. A further division by the interval length Δt yields time-dependent firing rate r(t) of the neuron, which is equivalent to the spike density of PSTH.

For sufficiently small Δt, r(t)Δt is the average number of spikes occurring between times t and t+Δt over multiple trials. If Δt is small, there will never be more than one spike within the interval between t and t+Δt on any given trial. This means that r(t)Δt is also the fraction of trials on which a spike occurred between those times. Equivalently, r(t)Δt is the probability that a spike occurs during this time interval.

As an experimental procedure, the time-dependent firing rate measure is a useful method to evaluate neuronal activity, in particular in the case of time-dependent stimuli. The obvious problem with this approach is that it can not be the coding scheme used by neurons in the brain. Neurons can not wait for the stimuli to repeatedly present in an exactly same manner before generating response.

Nevertheless, the experimental time-dependent firing rate measure can make sense, if there are large populations of independent neurons that receive the same stimulus. Instead of recording from a population of N neurons in a single run, it is experimentally easier to record from a single neuron and average over N repeated runs. Thus, the time-dependent firing rate coding relies on the implicit assumption that there are always populations of neurons.

Temporal coding

When precise spike timing or high-frequency firing-rate fluctuations are found to carry information, the neural code is often identified as a temporal code. A number of studies have found that the temporal resolution of the neural code is on a millisecond time scale, indicating that precise spike timing is a significant element in neural coding. Such codes, that communicate via the time between spikes are referred to as interpulse interval codes, and have been supported by recent studies.

Neurons exhibit high-frequency fluctuations of firing-rates which could be noise or could carry information. Rate coding models suggest that these irregularities are noise, while temporal coding models suggest that they encode information. If the nervous system only used rate codes to convey information, a more consistent, regular firing rate would have been evolutionarily advantageous, and neurons would have utilized this code over other less robust options. Temporal coding supplies an alternate explanation for the “noise," suggesting that it actually encodes information and affects neural processing. To model this idea, binary symbols can be used to mark the spikes: 1 for a spike, 0 for no spike. Temporal coding allows the sequence 000111000111 to mean something different from 001100110011, even though the mean firing rate is the same for both sequences, at 6 spikes/10 ms. Until recently, scientists had put the most emphasis on rate encoding as an explanation for post-synaptic potential patterns. However, functions of the brain are more temporally precise than the use of only rate encoding seems to allow^{[citation needed]}. In other words, essential information could be lost due to the inability of the rate code to capture all the available information of the spike train. In addition, responses are different enough between similar (but not identical) stimuli to suggest that the distinct patterns of spikes contain a higher volume of information than is possible to include in a rate code.

Temporal codes employ those features of the spiking activity that cannot be described by the firing rate. For example, time to first spike after the stimulus onset, characteristics based on the second and higher statistical moments of the ISI probability distribution, spike randomness, or precisely timed groups of spikes (temporal patterns) are candidates for temporal codes. As there is no absolute time reference in the nervous system, the information is carried either in terms of the relative timing of spikes in a population of neurons or with respect to an ongoing brain oscillation. One way in which temporal codes are decoded, in presence of neural oscillations, is that spikes occurring at specific phases of an oscillatory cycle are more effective in depolarizing the post-synaptic neuron.

The temporal structure of a spike train or firing rate evoked by a stimulus is determined both by the dynamics of the stimulus and by the nature of the neural encoding process. Stimuli that change rapidly tend to generate precisely timed spikes and rapidly changing firing rates no matter what neural coding strategy is being used. Temporal coding refers to temporal precision in the response that does not arise solely from the dynamics of the stimulus, but that nevertheless relates to properties of the stimulus. The interplay between stimulus and encoding dynamics makes the identification of a temporal code difficult.

In temporal coding, learning can be explained by activity-dependent synaptic delay modifications. The modifications can themselves depend not only on spike rates (rate coding) but also on spike timing patterns (temporal coding), i.e., can be a special case of spike-timing-dependent plasticity.

The issue of temporal coding is distinct and independent from the issue of independent-spike coding. If each spike is independent of all the other spikes in the train, the temporal character of the neural code is determined by the behavior of time-dependent firing rate r(t). If r(t) varies slowly with time, the code is typically called a rate code, and if it varies rapidly, the code is called temporal.

Temporal coding in sensory systems

For very brief stimuli, a neuron's maximum firing rate may not be fast enough to produce more than a single spike. Due to the density of information about the abbreviated stimulus contained in this single spike, it would seem that the timing of the spike itself would have to convey more information than simply the average frequency of action potentials over a given period of time. This model is especially important for sound localization, which occurs within the brain on the order of milliseconds. The brain must obtain a large quantity of information based on a relatively short neural response. Additionally, if low firing rates on the order of ten spikes per second must be distinguished from arbitrarily close rate coding for different stimuli, then a neuron trying to discriminate these two stimuli may need to wait for a second or more to accumulate enough information. This is not consistent with numerous organisms which are able to discriminate between stimuli in the time frame of milliseconds, suggesting that a rate code is not the only model at work.

To account for the fast encoding of visual stimuli, it has been suggested that neurons of the retina encode visual information in the latency time between stimulus onset and first action potential, also called latency to first spike. This type of temporal coding has been shown also in the auditory and somato-sensory system. The main drawback of such a coding scheme is its sensitivity to intrinsic neuronal fluctuations. In the primary visual cortex of macaques, the timing of the first spike relative to the start of the stimulus was found to provide more information than the interval between spikes. However, the interspike interval could be used to encode additional information, which is especially important when the spike rate reaches its limit, as in high-contrast situations. For this reason, temporal coding may play a part in coding defined edges rather than gradual transitions.

The mammalian gustatory system is useful for studying temporal coding because of its fairly distinct stimuli and the easily discernible responses of the organism. Temporally encoded information may help an organism discriminate between different tastants of the same category (sweet, bitter, sour, salty, umami) that elicit very similar responses in terms of spike count. The temporal component of the pattern elicited by each tastant may be used to determine its identity (e.g., the difference between two bitter tastants, such as quinine and denatonium). In this way, both rate coding and temporal coding may be used in the gustatory system – rate for basic tastant type, temporal for more specific differentiation. Research on mammalian gustatory system has shown that there is an abundance of information present in temporal patterns across populations of neurons, and this information is different from that which is determined by rate coding schemes. Groups of neurons may synchronize in response to a stimulus. In studies dealing with the front cortical portion of the brain in primates, precise patterns with short time scales only a few milliseconds in length were found across small populations of neurons which correlated with certain information processing behaviors. However, little information could be determined from the patterns; one possible theory is they represented the higher-order processing taking place in the brain.

As with the visual system, in mitral/tufted cells in the olfactory bulb of mice, first-spike latency relative to the start of a sniffing action seemed to encode much of the information about an odor. This strategy of using spike latency allows for rapid identification of and reaction to an odorant. In addition, some mitral/tufted cells have specific firing patterns for given odorants. This type of extra information could help in recognizing a certain odor, but is not completely necessary, as average spike count over the course of the animal's sniffing was also a good identifier. Along the same lines, experiments done with the olfactory system of rabbits showed distinct patterns which correlated with different subsets of odorants, and a similar result was obtained in experiments with the locust olfactory system.

Temporal coding applications

The specificity of temporal coding requires highly refined technology to measure informative, reliable, experimental data. Advances made in optogenetics allow neurologists to control spikes in individual neurons, offering electrical and spatial single-cell resolution. For example, blue light causes the light-gated ion channel channelrhodopsin to open, depolarizing the cell and producing a spike. When blue light is not sensed by the cell, the channel closes, and the neuron ceases to spike. The pattern of the spikes matches the pattern of the blue light stimuli. By inserting channelrhodopsin gene sequences into mouse DNA, researchers can control spikes and therefore certain behaviors of the mouse (e.g., making the mouse turn left). Researchers, through optogenetics, have the tools to effect different temporal codes in a neuron while maintaining the same mean firing rate, and thereby can test whether or not temporal coding occurs in specific neural circuits.

Optogenetic technology also has the potential to enable the correction of spike abnormalities at the root of several neurological and psychological disorders. If neurons do encode information in individual spike timing patterns, key signals could be missed by attempting to crack the code while looking only at mean firing rates. Understanding any temporally encoded aspects of the neural code and replicating these sequences in neurons could allow for greater control and treatment of neurological disorders such as depression, schizophrenia, and Parkinson's disease. Regulation of spike intervals in single cells more precisely controls brain activity than the addition of pharmacological agents intravenously.

Phase-of-firing code

Phase-of-firing code is a neural coding scheme that combines the spike count code with a time reference based on oscillations. This type of code takes into account a time label for each spike according to a time reference based on phase of local ongoing oscillations at low or high frequencies.

It has been shown that neurons in some cortical sensory areas encode rich naturalistic stimuli in terms of their spike times relative to the phase of ongoing network oscillatory fluctuations, rather than only in terms of their spike count. The local field potential signals reflect population (network) oscillations. The phase-of-firing code is often categorized as a temporal code although the time label used for spikes (i.e. the network oscillation phase) is a low-resolution (coarse-grained) reference for time. As a result, often only four discrete values for the phase are enough to represent all the information content in this kind of code with respect to the phase of oscillations in low frequencies. Phase-of-firing code is loosely based on the phase precession phenomena observed in place cells of the hippocampus. Another feature of this code is that neurons adhere to a preferred order of spiking between a group of sensory neurons, resulting in firing sequence.

Phase code has been shown in visual cortex to involve also high-frequency oscillations. Within a cycle of gamma oscillation, each neuron has its own preferred relative firing time. As a result, an entire population of neurons generates a firing sequence that has a duration of up to about 15 ms.

Population coding

Population coding is a method to represent stimuli by using the joint activities of a number of neurons. In population coding, each neuron has a distribution of responses over some set of inputs, and the responses of many neurons may be combined to determine some value about the inputs.
From the theoretical point of view, population coding is one of a few mathematically well-formulated problems in neuroscience. It grasps the essential features of neural coding and yet is simple enough for theoretic analysis. Experimental studies have revealed that this coding paradigm is widely used in the sensor and motor areas of the brain. For example, in the visual area medial temporal (MT), neurons are tuned to the moving direction. In response to an object moving in a particular direction, many neurons in MT fire with a noise-corrupted and bell-shaped activity pattern across the population. The moving direction of the object is retrieved from the population activity, to be immune from the fluctuation existing in a single neuron’s signal. In one classic example in the primary motor cortex, Apostolos Georgopoulos and colleagues trained monkeys to move a joystick towards a lit target. They found that a single neuron would fire for multiple target directions. However it would fire fastest for one direction and more slowly depending on how close the target was to the neuron's 'preferred' direction.

Kenneth Johnson originally derived that if each neuron represents movement in its preferred direction, and the vector sum of all neurons is calculated (each neuron has a firing rate and a preferred direction), the sum points in the direction of motion. In this manner, the population of neurons codes the signal for the motion. This particular population code is referred to as population vector coding. This particular study divided the field of motor physiologists between Evarts' "upper motor neuron" group, which followed the hypothesis that motor cortex neurons contributed to control of single muscles, and the Georgopoulos group studying the representation of movement directions in cortex.

The Johns Hopkins University Neural Encoding laboratory led by Murray Sachs and Eric Young developed place-time population codes, termed the Averaged-Localized-Synchronized-Response (ALSR) code for neural representation of auditory acoustic stimuli. This exploits both the place or tuning within the auditory nerve, as well as the phase-locking within each nerve fiber Auditory nerve. The first ALSR representation was for steady-state vowels; ALSR representations of pitch and formant frequencies in complex, non-steady state stimuli were demonstrated for voiced-pitch and formant representations in consonant-vowel syllables. The advantage of such representations is that global features such as pitch or formant transition profiles can be represented as global features across the entire nerve simultaneously via both rate and place coding.

Population coding has a number of other advantages as well, including reduction of uncertainty due to neuronal variability and the ability to represent a number of different stimulus attributes simultaneously. Population coding is also much faster than rate coding and can reflect changes in the stimulus conditions nearly instantaneously. Individual neurons in such a population typically have different but overlapping selectivities, so that many neurons, but not necessarily all, respond to a given stimulus.

Typically an encoding function has a peak value such that activity of the neuron is greatest if the perceptual value is close to the peak value, and becomes reduced accordingly for values less close to the peak value.

It follows that the actual perceived value can be reconstructed from the overall pattern of activity in the set of neurons. The Johnson/Georgopoulos vector coding is an example of simple averaging. A more sophisticated mathematical technique for performing such a reconstruction is the method of maximum likelihood based on a multivariate distribution of the neuronal responses. These models can assume independence, second order correlations, or even more detailed dependencies such as higher order maximum entropy models or copulas.

Correlation coding

The correlation coding model of neuronal firing claims that correlations between action potentials, or "spikes", within a spike train may carry additional information above and beyond the simple timing of the spikes. Early work suggested that correlation between spike trains can only reduce, and never increase, the total mutual information present in the two spike trains about a stimulus feature. However, this was later demonstrated to be incorrect. Correlation structure can increase information content if noise and signal correlations are of opposite sign. Correlations can also carry information not present in the average firing rate of two pairs of neurons. A good example of this exists in the pentobarbital-anesthetized marmoset auditory cortex, in which a pure tone causes an increase in the number of correlated spikes, but not an increase in the mean firing rate, of pairs of neurons.

Independent-spike coding

The independent-spike coding model of neuronal firing claims that each individual action potential, or "spike", is independent of each other spike within the spike train.

Position coding

Plot of typical position coding

A typical population code involves neurons with a Gaussian tuning curve whose means vary linearly with the stimulus intensity, meaning that the neuron responds most strongly (in terms of spikes per second) to a stimulus near the mean. The actual intensity could be recovered as the stimulus level corresponding to the mean of the neuron with the greatest response. However, the noise inherent in neural responses means that a maximum likelihood estimation function is more accurate.

Neural responses are noisy and unreliable.

This type of code is used to encode continuous variables such as joint position, eye position, color, or sound frequency. Any individual neuron is too noisy to faithfully encode the variable using rate coding, but an entire population ensures greater fidelity and precision. For a population of unimodal tuning curves, i.e. with a single peak, the precision typically scales linearly with the number of neurons. Hence, for half the precision, half as many neurons are required. In contrast, when the tuning curves have multiple peaks, as in grid cells that represent space, the precision of the population can scale exponentially with the number of neurons. This greatly reduces the number of neurons required for the same precision.

Sparse coding

The sparse code is when each item is encoded by the strong activation of a relatively small set of neurons. For each item to be encoded, this is a different subset of all available neurons. In contrast to sensor-sparse coding, sensor-dense coding implies that all information from possible sensor locations is known.

As a consequence, sparseness may be focused on temporal sparseness ("a relatively small number of time periods are active") or on the sparseness in an activated population of neurons. In this latter case, this may be defined in one time period as the number of activated neurons relative to the total number of neurons in the population. This seems to be a hallmark of neural computations since compared to traditional computers, information is massively distributed across neurons. A major result in neural coding from Olshausen and Field is that sparse coding of natural images produces wavelet-like oriented filters that resemble the receptive fields of simple cells in the visual cortex. The capacity of sparse codes may be increased by simultaneous use of temporal coding, as found in the locust olfactory system.

Given a potentially large set of input patterns, sparse coding algorithms (e.g. Sparse Autoencoder) attempt to automatically find a small number of representative patterns which, when combined in the right proportions, reproduce the original input patterns. The sparse coding for the input then consists of those representative patterns. For example, the very large set of English sentences can be encoded by a small number of symbols (i.e. letters, numbers, punctuation, and spaces) combined in a particular order for a particular sentence, and so a sparse coding for English would be those symbols.

Linear generative model

Most models of sparse coding are based on the linear generative model. In this model, the symbols are combined in a linear fashion to approximate the input.

More formally, given a k-dimensional set of real-numbered input vectors ${\displaystyle {\vec {\xi }}\in \mathbb {R} ^{k}}$, the goal of sparse coding is to determine n k-dimensional basis vectors ${\displaystyle {\vec {b_{1}}},\ldots ,{\vec {b_{n}}}\in \mathbb {R} ^{k}}$ along with a sparse n-dimensional vector of weights or coefficients ${\displaystyle {\vec {s}}\in \mathbb {R} ^{n}}$ for each input vector, so that a linear combination of the basis vectors with proportions given by the coefficients results in a close approximation to the input vector: ${\displaystyle {\vec {\xi }}\approx \sum _{j=1}^{n}s_{j}{\vec {b}}_{j}}$.

The codings generated by algorithms implementing a linear generative model can be classified into codings with soft sparseness and those with hard sparseness. These refer to the distribution of basis vector coefficients for typical inputs. A coding with soft sparseness has a smooth Gaussian-like distribution, but peakier than Gaussian, with many zero values, some small absolute values, fewer larger absolute values, and very few very large absolute values. Thus, many of the basis vectors are active. Hard sparseness, on the other hand, indicates that there are many zero values, no or hardly any small absolute values, fewer larger absolute values, and very few very large absolute values, and thus few of the basis vectors are active. This is appealing from a metabolic perspective: less energy is used when fewer neurons are firing.

Another measure of coding is whether it is critically complete or overcomplete. If the number of basis vectors n is equal to the dimensionality k of the input set, the coding is said to be critically complete. In this case, smooth changes in the input vector result in abrupt changes in the coefficients, and the coding is not able to gracefully handle small scalings, small translations, or noise in the inputs. If, however, the number of basis vectors is larger than the dimensionality of the input set, the coding is overcomplete. Overcomplete codings smoothly interpolate between input vectors and are robust under input noise. The human primary visual cortex is estimated to be overcomplete by a factor of 500, so that, for example, a 14 x 14 patch of input (a 196-dimensional space) is coded by roughly 100,000 neurons.

Biological evidence

Sparse coding may be a general strategy of neural systems to augment memory capacity. To adapt to their environments, animals must learn which stimuli are associated with rewards or punishments and distinguish these reinforced stimuli from similar but irrelevant ones. Such task requires implementing stimulus-specific associative memories in which only a few neurons out of a population respond to any given stimulus and each neuron responds to only a few stimuli out of all possible stimuli.
Theoretical work on Sparse distributed memory has suggested that sparse coding increases the capacity of associative memory by reducing overlap between representations. Experimentally, sparse representations of sensory information have been observed in many systems, including vision, audition, touch, and olfaction. However, despite the accumulating evidence for widespread sparse coding and theoretical arguments for its importance, a demonstration that sparse coding improves the stimulus-specificity of associative memory has been lacking until recently.

Some progress has been made in 2014 by Gero Miesenböck's lab at the University of Oxford analyzing Drosophila Olfactory system. In Drosophila, sparse odor coding by the Kenyon cells of the mushroom body is thought to generate a large number of precisely addressable locations for the storage of odor-specific memories. Lin et al. demonstrated that sparseness is controlled by a negative feedback circuit between Kenyon cells and the GABAergic anterior paired lateral (APL) neuron. Systematic activation and blockade of each leg of this feedback circuit show that Kenyon cells activate APL and APL inhibits Kenyon cells. Disrupting the Kenyon cell-APL feedback loop decreases the sparseness of Kenyon cell odor responses, increases inter-odor correlations, and prevents flies from learning to discriminate similar, but not dissimilar, odors. These results suggest that feedback inhibition suppresses Kenyon cell activity to maintain sparse, decorrelated odor coding and thus the odor-specificity of memories.

Swarm intelligence

From Wikipedia, the free encyclopedia

 

Swarm intelligence (SI) is the collective behavior of decentralized, self-organized systems, natural or artificial. The concept is employed in work on artificial intelligence. The expression was introduced by Gerardo Beni and Jing Wang in 1989, in the context of cellular robotic systems.

 

SI systems consist typically of a population of simple agents or boids interacting locally with one another and with their environment. The inspiration often comes from nature, especially biological systems. The agents follow very simple rules, and although there is no centralized control structure dictating how individual agents should behave, local, and to a certain degree random, interactions between such agents lead to the emergence of "intelligent" global behavior, unknown to the individual agents. Examples in natural systems of SI include ant colonies, bird flocking, animal herding, bacterial growth, fish schooling and microbial intelligence.

The application of swarm principles to robots is called swarm robotics, while 'swarm intelligence' refers to the more general set of algorithms. 'Swarm prediction' has been used in the context of forecasting problems.

Models of swarm behavior

Boids (Reynolds 1987)

Boids is an artificial life program, developed by Craig Reynolds in 1986, which simulates the flocking behaviour of birds. His paper on this topic was published in 1987 in the proceedings of the ACM SIGGRAPH conference. The name "boid" corresponds to a shortened version of "bird-oid object", which refers to a bird-like object.

As with most artificial life simulations, Boids is an example of emergent behavior; that is, the complexity of Boids arises from the interaction of individual agents (the boids, in this case) adhering to a set of simple rules. The rules applied in the simplest Boids world are as follows:

• separation: steer to avoid crowding local flockmates
• alignment: steer towards the average heading of local flockmates
• cohesion: steer to move toward the average position (center of mass) of local flockmates

More complex rules can be added, such as obstacle avoidance and goal seeking.

Self-propelled particles (Vicsek et al. 1995)

Self-propelled particles (SPP), also referred to as the Vicsek model, was introduced in 1995 by Vicsek et al. as a special case of the boids model introduced in 1986 by Reynolds. A swarm is modelled in SPP by a collection of particles that move with a constant speed but respond to a random perturbation by adopting at each time increment the average direction of motion of the other particles in their local neighbourhood. SPP models predict that swarming animals share certain properties at the group level, regardless of the type of animals in the swarm. Swarming systems give rise to emergent behaviours which occur at many different scales, some of which are turning out to be both universal and robust. It has become a challenge in theoretical physics to find minimal statistical models that capture these behaviours.

Metaheuristics

Evolutionary algorithms (EA), particle swarm optimization (PSO), ant colony optimization (ACO) and their variants dominate the field of nature-inspired metaheuristics. This list includes algorithms published up to circa the year 2000. A large number of more recent metaphor-inspired metaheuristics have started to attract criticism in the research community for hiding their lack of novelty behind an elaborate metaphor. For algorithms published since that time, see List of metaphor-based metaheuristics.

Stochastic diffusion search (Bishop 1989)

First published in 1989 Stochastic diffusion search (SDS) was the first Swarm Intelligence metaheuristic. SDS is an agent-based probabilistic global search and optimization technique best suited to problems where the objective function can be decomposed into multiple independent partial-functions. Each agent maintains a hypothesis which is iteratively tested by evaluating a randomly selected partial objective function parameterised by the agent's current hypothesis. In the standard version of SDS such partial function evaluations are binary, resulting in each agent becoming active or inactive. Information on hypotheses is diffused across the population via inter-agent communication. Unlike the stigmergic communication used in ACO, in SDS agents communicate hypotheses via a one-to-one communication strategy analogous to the tandem running procedure observed in Leptothorax acervorum. A positive feedback mechanism ensures that, over time, a population of agents stabilise around the global-best solution. SDS is both an efficient and robust global search and optimisation algorithm, which has been extensively mathematically described. Recent work has involved merging the global search properties of SDS with other swarm intelligence algorithms.

Ant colony optimization (Dorigo 1992)

Ant colony optimization (ACO), introduced by Dorigo in his doctoral dissertation, is a class of optimization algorithms modeled on the actions of an ant colony. ACO is a probabilistic technique useful in problems that deal with finding better paths through graphs. Artificial 'ants'—simulation agents—locate optimal solutions by moving through a parameter space representing all possible solutions. Natural ants lay down pheromones directing each other to resources while exploring their environment. The simulated 'ants' similarly record their positions and the quality of their solutions, so that in later simulation iterations more ants locate for better solutions.

Particle swarm optimization (Kennedy, Eberhart & Shi 1995)

Particle swarm optimization (PSO) is a global optimization algorithm for dealing with problems in which a best solution can be represented as a point or surface in an n-dimensional space. Hypotheses are plotted in this space and seeded with an initial velocity, as well as a communication channel between the particles. Particles then move through the solution space, and are evaluated according to some fitness criterion after each timestep. Over time, particles are accelerated towards those particles within their communication grouping which have better fitness values. The main advantage of such an approach over other global minimization strategies such as simulated annealing is that the large number of members that make up the particle swarm make the technique impressively resilient to the problem of local minima.

Applications

Swarm Intelligence-based techniques can be used in a number of applications. The U.S. military is investigating swarm techniques for controlling unmanned vehicles. The European Space Agency is thinking about an orbital swarm for self-assembly and interferometry. NASA is investigating the use of swarm technology for planetary mapping. A 1992 paper by M. Anthony Lewis and George A. Bekey discusses the possibility of using swarm intelligence to control nanobots within the body for the purpose of killing cancer tumors. Conversely al-Rifaie and Aber have used stochastic diffusion search to help locate tumours. Swarm intelligence has also been applied for data mining.

Ant-based routing

The use of swarm intelligence in telecommunication networks has also been researched, in the form of ant-based routing. This was pioneered separately by Dorigo et al. and Hewlett Packard in the mid-1990s, with a number of variations since. Basically, this uses a probabilistic routing table rewarding/reinforcing the route successfully traversed by each "ant" (a small control packet) which flood the network. Reinforcement of the route in the forwards, reverse direction and both simultaneously have been researched: backwards reinforcement requires a symmetric network and couples the two directions together; forwards reinforcement rewards a route before the outcome is known (but then one would pay for the cinema before one knows how good the film is). As the system behaves stochastically and is therefore lacking repeatability, there are large hurdles to commercial deployment. Mobile media and new technologies have the potential to change the threshold for collective action due to swarm intelligence (Rheingold: 2002, P175).

The location of transmission infrastructure for wireless communication networks is an important engineering problem involving competing objectives. A minimal selection of locations (or sites) are required subject to providing adequate area coverage for users. A very different-ant inspired swarm intelligence algorithm, stochastic diffusion search (SDS), has been successfully used to provide a general model for this problem, related to circle packing and set covering. It has been shown that the SDS can be applied to identify suitable solutions even for large problem instances.

Airlines have also used ant-based routing in assigning aircraft arrivals to airport gates. At Southwest Airlines a software program uses swarm theory, or swarm intelligence—the idea that a colony of ants works better than one alone. Each pilot acts like an ant searching for the best airport gate. "The pilot learns from his experience what's the best for him, and it turns out that that's the best solution for the airline," Douglas A. Lawson explains. As a result, the "colony" of pilots always go to gates they can arrive at and depart from quickly. The program can even alert a pilot of plane back-ups before they happen. "We can anticipate that it's going to happen, so we'll have a gate available," Lawson says.

Crowd simulation

Artists are using swarm technology as a means of creating complex interactive systems or simulating crowds.

Stanley and Stella in: Breaking the Ice was the first movie to make use of swarm technology for rendering, realistically depicting the movements of groups of fish and birds using the Boids system. Tim Burton's Batman Returns also made use of swarm technology for showing the movements of a group of bats. The Lord of the Rings film trilogy made use of similar technology, known as Massive, during battle scenes. Swarm technology is particularly attractive because it is cheap, robust, and simple.

Airlines have used swarm theory to simulate passengers boarding a plane. Southwest Airlines researcher Douglas A. Lawson used an ant-based computer simulation employing only six interaction rules to evaluate boarding times using various boarding methods.(Miller, 2010, xii-xviii).

Human swarming

Enabled by mediating software such as the SWARM platform (formally unu) from Unanimous A.I., networks of distributed users can be organized into "human swarms" through the implementation of real-time closed-loop control systems. As published by Rosenberg (2015), such real-time systems enable groups of human participants to behave as a unified collective intelligence that works as a single entity to make predictions, answer questions, and evoke opinions. Such systems, also referred to as "Artificial Swarm Intelligence" (or the brand name Swarm AI) have been shown to significantly amplify human intelligence, resulting in a string of high-profile predictions of extreme accuracy. Academic testing shows that human swarms can out-predict individuals across a variety of real-world projections. Famously, human swarming was used to correctly predict the Kentucky Derby Superfecta, against 541 to 1 odds, in response to a challenge from reporters.

Swarm grammars

Swarm grammars are swarms of stochastic grammars that can be evolved to describe complex properties such as found in art and architecture. These grammars interact as agents behaving according to rules of swarm intelligence. Such behavior can also suggest deep learning algorithms, in particular when mapping of such swarms to neural circuits is considered.

Swarmic art

In a series of works al-Rifaie et al. have successfully used two swarm intelligence algorithms—one mimicking the behaviour of one species of ants (Leptothorax acervorum) foraging (stochastic diffusion search, SDS) and the other algorithm mimicking the behaviour of birds flocking (particle swarm optimization, PSO)—to describe a novel integration strategy exploiting the local search properties of the PSO with global SDS behaviour. The resulting hybrid algorithm is used to sketch novel drawings of an input image, exploiting an artistic tension between the local behaviour of the 'birds flocking'—as they seek to follow the input sketch—and the global behaviour of the "ants foraging"—as they seek to encourage the flock to explore novel regions of the canvas. The "creativity" of this hybrid swarm system has been analysed under the philosophical light of the "rhizome" in the context of Deleuze's "Orchid and Wasp" metaphor.

In a more recent work of al-Rifaie et al., "Swarmic Sketches and Attention Mechanism", introduces a novel approach deploying the mechanism of 'attention' by adapting SDS to selectively attend to detailed areas of a digital canvas. Once the attention of the swarm is drawn to a certain line within the canvas, the capability of PSO is used to produce a 'swarmic sketch' of the attended line. The swarms move throughout the digital canvas in an attempt to satisfy their dynamic roles—attention to areas with more details—associated to them via their fitness function. Having associated the rendering process with the concepts of attention, the performance of the participating swarms creates a unique, non-identical sketch each time the 'artist' swarms embark on interpreting the input line drawings. In other works while PSO is responsible for the sketching process, SDS controls the attention of the swarm.

In a similar work, "Swarmic Paintings and Colour Attention", non-photorealistic images are produced using SDS algorithm which, in the context of this work, is responsible for colour attention.
The "computational creativity" of the above-mentioned systems are discussed in through the two prerequisites of creativity (i.e. freedom and constraints) within the swarm intelligence's two infamous phases of exploration and exploitation.

Michael Theodore and Nikolaus Correll use swarm intelligent art installation to explore what it takes to have engineered systems to appear lifelike Notable work include Swarm Wall (2012) and endo-exo (2014).

Distributed cognition

From Wikipedia, the free encyclopedia

 

Distributed cognition is an approach to cognitive science research that deploys models of the extended mind by taking as the fundamental unit of analysis "a collection of individuals and artifacts and their relations to each other in a particular work practice". "DCog" is a specific approach to distributed cognition (distinct from other meanings) which takes a computational perspective towards goal-based activity systems. Dcog frameworks employed were originally developed in the mid-1980s by Edwin Hutchins, who continues to be the leading pioneer and whose research is based at the University of California at San Diego.

 

Using insights from sociology, cognitive science, and the psychology of Vygotsky (cf. cultural-historical psychology) it emphasizes the ways that cognition is off-loaded into the environment through social and technological means. It is a framework for studying cognition rather than a type of cognition. This framework involves the coordination between individuals, artifacts and the environment. According to Zhang & Norman (1994), the distributed cognition approach has three key components:

1. Embodiment of information that is embedded in representations of interaction
2. Coordination of enaction among embodied agents
3. Ecological contributions to a cognitive ecosystem

'Dcog' studies the "propagation of representational states across media" (Rogers and Ellis, ibid.). Mental content is considered to be non-reducible to individual cognition and is more properly understood as off-loaded and extended into the environment, where information is also made available to other agents (Heylighen, Heath, & Overwalle, 2003). It is often understood as an approach in specific opposition to earlier and still prevalent "brain in a vat" models which ignore "situatedness, embodiment and enaction" as key to any cognitive act (Ibid.).

These representation-based frameworks consider distributed cognition as "a cognitive system whose structures and processes are distributed between internal and external representations, across a group of individuals, and across space and time" (Zhang and Patel, 2006). In general terms, they consider a distributed cognition system to have two components: internal and external representations. In their description, internal representations are knowledge and structure in individuals' minds while external representations are knowledge and structure in the external environment (Zhang, 1997b; Zhang and Norman, 1994).

DCog studies the ways that memories, facts, or knowledge is embedded in the objects, individuals, and tools in our environment. DCog is a useful approach for (re)designing the technologically mediated social aspects of cognition by putting emphasis on the individual and his/her environment, and the media channels with which people interact, either in order to communicate with each other, or socially coordinate to perform complex tasks. Distributed cognition views a system of cognition as a set of representations propagated through specific media, and models the interchange of information between these representational media. These representations can be either in the mental space of the participants or external representations available in the environment.

These interactions can be categorized into three distinct types of processes:

1. Cognitive processes may be distributed across the members of a social group.
2. Cognitive processes may be distributed in the sense that the operation of the cognitive system involves coordination between internal and external (material or environmental) structure.
3. Processes may be distributed through time in such a way that the products of earlier events can transform the nature of related events.

Early research

John Milton Roberts thought that social organization could be seen as cognition through a community (Roberts 1964). He described the cognitive aspects of a society by looking at the present information and how it moves through the people in the society.

Daniel L. Schwartz (1978) proposed a distribution of cognition through culture and the distribution of beliefs across the members of a society.

In 1998, Mark Perry from Brunel University London explored the problems and the benefits brought by distributed cognition to "understanding the organisation of information within its contexts." He considered that distributed cognition draws from the information processing metaphor of cognitive science where a system is considered in terms of its inputs and outputs and tasks are decomposed into a problem space (Perry, 1998). He believed that information should be studied through the representation within the media or artifact that represents the information. Cognition is said to be "socially distributed" when it is applied to demonstrate how interpersonal processes can be used to coordinate activity within a social group.

In 1999, Gavriel Salomon stated that there were two classes of distributive cognition: shared cognition and off-loading. Shared cognition is that which is shared among people through common activity such as conversation where there is a constant change of cognition based on the other person's responses. An example of off-loading would be using a calculator to do arithmetic or a creating a grocery list when going shopping. In that sense, the cognitive duties are off-loaded to a material object.

Later, John Sutton (2006) defined five appropriate domains of investigation for research in Dcog:

1. External cultural tools, artifacts, and symbol systems.
2. Natural environmental resources.
3. Interpersonal and social distribution or scaffolding.
4. Embodied capacities and skills.
5. Internalized cognitive artifacts.

Applications

The application area of DCog is systems design and implementation in specific work environments. Its main method is field research, going into the workplace and making rigorous observations, e.g. through capturing work performances with video, studying and coding the recorded activities using qualitative research methods to codify the various ways in which cognition is distributed in the local environment, through the social and technical systems with which the workers engage.

Distributed cognition as a theory of learning, i.e. one in which the development of knowledge is attributed to the system of thinking agents interacting dynamically with artifacts, has been widely applied in the field of distance learning, especially in relation to computer-supported collaborative learning (CSCL) and other computer-supported learning tools. For example, in the field of teaching English Composition, Kevin LaGrandeur has argued that CSCL provides a source of common memory, collaborative space, and a cognitive artifact (tool to enhance cognition) that allows students to more easily build effective written compositions via explicit and implicit machine-human collaboration. Distributed cognition illustrates the process of interaction between people and technologies in order to determine how to best represent, store and provide access to digital resources and other artifacts.

Collaborative tagging on the World Wide Web is one of the most recent developments in technological support for distributed cognition. Beginning in 2004 and quickly becoming a standard on websites, collaborative tagging allows users to upload or select materials (e.g. pictures, music files, texts, websites) and associate tags with these materials. Tags can be chosen freely, and are similar to keywords. Other users can then browse through tags; a click on a tag connects a user to similarly tagged materials. Tags furthermore enable tag clouds, which graphically represent the popularity of tags, demonstrating co-occurrence relations between tags and thus jump from one tag to another.

Dcog has also been used to understand learning and communication in clinical settings and to obtain an integrated view of clinical workplace learning. It has been observed how medical actors use and connect gestural practices, along with visual and haptic structures of their own bodies and of artifacts such as technological instruments and computational devices. In so doing they co-construct complex, multimodal representations that go beyond the mental representations usually studied from a cognitive perspective of learning (Pimmer, Pachler & Genewein, 2013).

Distributed cognition can also be seen through cultures and communities. Learning certain habits or following certain traditions is seen as cognition distributed over a group of people. Exploring distributed cognition through community and culture is one way to understand how it may work.
With the new research that is emerging in this field, the overarching concept of distributed cognition enhances the understanding of interactions between individual human beings and artifacts such as technologies and machines, and complex external environments. This concept has been applied to educational research in the areas of distributed leadership and distributed instruction.

Metaphors and examples

Distributed cognition is seen when using paper and pencil to do a complicated arithmetic problem. The person doing the problem may talk with a friend to clarify the problem, and then must write the partial answers on the paper in order to be able to keep track of all the steps in the calculation. In this example, the parts of distributed cognition are seen in:

• setting up the problem, in collaboration with another person,
• performing manipulation/arithmetic procedures, both in one's head and by writing down resulting partial answers.

The process of working out the answer requires not only the perception and thought of two people, it also requires the use of a tool (paper) to extend an individual's memory. So the intelligence is distributed, both between people, and a person and an object.

Another well-researched site for analyzing distributed cognition and applying the discovered insights towards the design of more optimal systems is aviation,where both cockpits and air traffic control environments have been studied as scenes that technologically and socially distribute cognition through systems of externalized representational media. It is not the cognitive performance and expertise of any one single person or machine that is important for the continued operation or the landing and takeoff of airplanes. The cognition is distributed over the personnel, sensors, and machinery both in the plane and on the ground, including but not limited to the controllers, pilots and crew as a whole.

Hutchins also examined another scene of distributed cognition within the context of navigating a US navy vessel. In his book on USS Palau, he explains in detail how distributed cognition is manifested through the interaction between crew members as they interpret, process, and transform information into various representational states in order to safely navigate the ship. In this functional unit, crew members (e.g. pelorus operators, bearing takers, plotters, and the ship's captain) play the role of actors who transform information into different representational states (i.e. triangulation, landmark sightings, bearings, and maps). In this context, navigation is embodied through the combined efforts of actors in the functional unit.

In his study on process, representation and taskworld, Mark Perry (1998) demonstrated how distributed cognition analysis can be conducted in a field study. His example was design analysis in Civil engineering. In this work, he showed how an information processing approach can be applied by carrying a detailed analysis of the background of the study - goals and resources, inputs and outputs, representations and processes, and transformational activity, "how information was transformed from the design drawings and site onto tables of measurements (different representations)" and then onto "a graphical representation" which provided a clearer demonstration of the relationship between the two data sets (Perry, 1998).

Quotes

On educational psychology:

People think in conjunction and partnership with others and with the help of culturally provided tools and implements.

— Salomon, 1997 p. xiii

On cognitive science:

Nervous systems do not form representations of the world, they can only form representations of interactions with the world.

The emphasis on finding and describing "knowledge structures" that are somewhere "inside" the individual encourages us to overlook the fact that human cognition is always situated in a complex sociocultural world and cannot be unaffected by it.

— Hutchins, 1995 p. xiii

Connectivism

From Wikipedia, the free encyclopedia

 

Connectivism is a theory of learning in a digital age that emphasizes the role of social and cultural context in how and where learning occurs. Learning does not simply happen within an individual, but within and across the networks. What sets connectivism apart from theories such as constructivism is the view that "learning (defined as actionable knowledge) can reside outside of ourselves (within an organization or a database), is focused on connecting specialized information sets, and the connections that enable us to learn more are more important than our current state of knowing". Connectivism sees knowledge as a network and learning as a process of pattern recognition. Connectivism has similarities with Vygotsky's 'zone of proximal development' (ZPD) and Engeström's Activity theory. The phrase "a learning theory for the digital age" indicates the emphasis that connectivism gives to technology's effect on how people live, communicate, and learn.

Nodes and links

The central aspect of connectivism is the metaphor of a network with nodes and connections. In this metaphor, a node is anything that can be connected to another node such as an organization, information, data, feelings, and images. Connectivism recognizes three node types: neural, conceptual (internal) and external. Connectivism sees learning as the process of creating connections and expanding or increasing network complexity. Connections may have different directions and strength. In this sense, a connection joining nodes A and B which goes from A to B is not the same as one that goes from B to A. There are some special kinds of connections such as "self-join" and pattern. A self-join connection joins a node to itself and a pattern can be defined as "a set of connections appearing together as a single whole".

The idea of organisation as cognitive systems where knowledge is distributed across nodes originated from the Perceptron (Artificial neuron) in an Artificial Neural Network, and is directly borrowed from Connectionism, "a software structure developed based on concepts inspired by biological functions of brain; it aims at creating machines able to learn like human".

The network metaphor allows a notion of "know-where" (the understanding of where to find the knowledge when it is needed) to supplement to the ones of "know-how" and "know-what" that make the cornerstones of many theories of learning.

As Downes states: "at its heart, connectivism is the thesis that knowledge is distributed across a network of connections, and therefore that learning consists of the ability to construct and traverse those networks".

Principles

• Learning and knowledge rests in diversity of opinions.
• Learning is a process of connecting specialized nodes or information sources.
• Learning may reside in non-human appliances.
• Learning is more critical than knowing.
• Maintaining and nurturing connections is needed to facilitate continual learning.
• Perceiving connections between fields, ideas and concepts is a core skill.
• Currency (accurate, up-to-date knowledge) is the intent of learning activities.
• Decision-making is itself a learning process. Choosing what to learn and the meaning of incoming information is seen through the lens of a shifting reality. While there is a right answer now, it may be wrong tomorrow due to alterations in the information climate affecting the decision.

Teaching methods

Summarizing connectivist teaching and learning, Downes states: "to teach is to model and demonstrate, to learn is to practice and reflect."

In 2008, Siemens and Downes delivered an online course called "Connectivism and Connective Knowledge". It covered connectivism as content while attempting to implement some of their ideas. The course was free to anyone who wished to participate, and over 2000 people worldwide enrolled. The phrase "Massive Open Online Course" (MOOC) describes this model. All course content was available through RSS feeds, and learners could participate with their choice of tools: threaded discussions in Moodle, blog posts, Second Life and synchronous online meetings. The course was repeated in 2009 and in 2011.

At its core, connectivism is a form of experiential learning which prioritizes the set of formed by actions and experience over the idea that knowledge is propositional.

History

Connectivism was introduced in 2005 by two publications, Siemens’ Connectivism: Learning as Network Creation and Downes’ An Introduction to Connective Knowledge. Both works received significant attention in the blogosphere and an extended discourse has followed on the appropriateness of connectivism as a learning theory for the digital age. In 2007 Kerr entered into the debate with a series of lectures and talks on the matter, as did Forster, both at the Online Connectivism Conference at the University of Manitoba. In 2008, in the context of digital and e-learning, connectivism was reconsidered and its technological implications were discussed by Siemens' and Ally.

Criticisms

The idea that connectivism is a new theory of learning is not widely accepted. Verhagen argued that connectivism is rather a "pedagogical view."

The lack of comparative literature reviews in Connectivism papers complicate evaluating how Connectivism relates to prior theories, such as Socially Distributed Cognition (Hutchins, 1995), which explored how connectionist ideas could be applied to social systems. Classical theories of cognition such as Activity theory (Vygotsky, Leont’ev, Luria, and others starting in the 1920s) proposed that people are embedded actors, with learning considered via three features – a subject (the learner), an object (the task or activity) and tool or mediating artifacts. Social cognitive theory (Bandura, 1962) claimed that people learn by watching others. Social learning theory (Miller and Dollard) elaborated this notion. Situated cognition (Brown, Collins, & Duguid, 1989; Greeno & Moore, 1993) alleged that knowledge is situated in activity bound to social, cultural and physical contexts; knowledge and learning that requires thinking on the fly rather than the storage and retrieval of conceptual knowledge. Community of practice (Lave & Wenger 1991) asserted that the process of sharing information and experiences with the group enables members to learn from each other. Collective intelligence (Lévy, 1994) described a shared or group intelligence that emerges from collaboration and competition.

Kerr claims that although technology affects learning environments, existing learning theories are sufficient. Kop and Hill conclude that while it does not seem that connectivism is a separate learning theory, it "continues to play an important role in the development and emergence of new pedagogies, where control is shifting from the tutor to an increasingly more autonomous learner."

AlDahdouh examined the relation between connectivism and Artificial Neural Network (ANN) and the results, unexpectedly, revealed that ANN researchers use constructivism principles to teach ANN with labeled training data. However, he argued that connectivism principles are used to teach ANN only when the knowledge is unknown.

Ally recognizes that the world has changed and become more networked, so learning theories developed prior to these global changes are less relevant. However, he argues that, "What is needed is not a new stand-alone theory for the digital age, but a model that integrates the different theories to guide the design of online learning materials.".

Chatti notes that Connectivism misses some concepts, which are crucial for learning, such as reflection, learning from failures, error detection and correction, and inquiry. He introduces the Learning as a Network (LaaN) theory which builds upon connectivism, complexity theory, and double-loop learning. LaaN starts from the learner and views learning as the continuous creation of a personal knowledge network (PKN).