Abstract
Attempting to imitate the brain’s functionalities, researchers have bridged between neuroscience and artificial intelligence for decades; however, experimental neuroscience has not directly advanced the field of machine learning (ML). Here, using neuronal cultures, we demonstrate that increased training frequency accelerates the neuronal adaptation processes. This mechanism was implemented on artificial neural networks, where a local learning stepsize increases for coherent consecutive learning steps, and tested on a simple dataset of handwritten digits, MNIST. Based on our online learning results with a few handwriting examples, success rates for braininspired algorithms substantially outperform the commonly used ML algorithms. We speculate this emerging bridge from slow brain function to ML will promote ultrafast decision making under limited examples, which is the reality in many aspects of human activity, robotic control, and network optimization.
Introduction
Machine learning is based on Donald Hebb’s pioneering work; seventy years ago, he suggested that learning occurs in the brain through synaptic (link) strength modifications^{1}. A synaptic strength modification typically lasts tens of minutes^{2} while the clock speed of a neuron (node) ranges around one second^{3}. Although the brain is comparatively slow, its computational capabilities outperform typical stateoftheart artificial intelligence algorithms. Following this speed/capability paradox, we experimentally derive accelerated learning mechanisms based on small datasets, where their utilization on gigahertz processors is expected to lead to ultrafast decision making.
Unlike modern computers, a welldefined global clock does not govern brain dynamics; instead, they are a function of relative event timing (e.g., stimulations and evoked spikes)^{4}. According to neuronal computational, using decaying input summation via its ramified dendritic trees, each neuron sums the asynchronous incoming electrical signals and generates a short electrical pulse (spike) when its threshold is reached. For each neuron, synaptic strength is slowly modified based on the relative timing of inputs from other synapses; if a signal is induced from a synapse without generating a spike, its associated strength is modified based on the relative timing to adjacent spikes from other synapses on the same neuron^{5}.
Recently it was experimentally demonstrated that each neuron functions as a collection of independent threshold units^{6}. After signals arrive via one of the dendritic trees, each threshold unit is activated. Additionally, a new type of adaptive rule was experimentally observed based on dendritic signal arrival timing^{7}, which is similar to the slow adaptation mechanism currently attributed to synapses (links). This dendritic adaptation occurs on a faster timescale: it requires approximately five minutes, while synaptic modification requires tens of minutes or more.
Results
In this study, dendritic adaptation was experimentally examined at a higher stimulation frequency, 5 Hz, using the training pattern of previous experiments run at 0.5 to 1 Hz. We planted neuronal cultures on a multielectrodearray with added synaptic blockers, which extracellularly stimulated a patched neuron via its dendrites (Fig. 1a and Materials and Methods). The adaptation process consisted of a training set: 50 pairs of stimulations. After an abovethreshold intracellular stimulation, an extracellular stimulation that did not evoke a spike arrived with a predefined delay, typically 1 to 4 ms (Fig. 1b). We primarily take into account differing extra and intraspike waveforms (Fig. 1a, right), which presumably activated the neuron from two independent dendritic trees^{7}.
By comparing the amplitudes of intracellular responses and extracellular stimulation before and after the training procedure, we quantified the effect of the neuronal adaptation. To quantify the initial response, the extracellular stimulation amplitude was decreased until no reliable evoked spikes were observed (Fig. 1c,d, left).
In the first type of experiment, one minute after the training terminated, we measured the enhanced responses and witnessed dendritic adaptation (Fig. 1c, right). When compared with the visible adaptation time for the 1 Hz training, that of the 5 Hz training was substantially faster. Occasionally, the visible effect of adaptation was further enhanced after more time passed (Supplementary Fig. 1), suggesting by extrapolation that adaptation might occur in much less than one minute. Because after training, the initialization and compilation of subsequent experiments required a minimal time lag (one minute), the feasibility of such ultrafast adaptation was impossible to examine.
To overcome this limitation, we introduced a second type of experiment (Fig. 1d): shortly after the end of the training procedure, the neuron was extracellularly stimulated using two predefined amplitudes with unreliable responses. Without requiring any new experimental methods or compilations (Materials and Methods), this procedure pinpointed dendritic adaptation within only 10 seconds of the training termination (Fig. 1d, right).
With the increased training frequency, the adaptation process substantially accelerated (Fig. 2a), potentially implying a timedependent decaying adaptation stepsize:
where the current adaptation step, \({\eta }_{adap}^{t+1}\), is equal to the previous one with a decaying weight, \(t\) stands for a discrete time step, \({\tau }_{0}\) is a constant, 1/τ stands for the training frequency, and Δ is a constant representing the incremental effect of the current training step. This type of decay process occurs in many biological scenarios and represents, for instance, the decaying concentration of active material due to diffusion. As a generalization of Eq. (1), incoherent consecutive training steps are also allowed, decreasing the dendritic strength and resulting in a −Δ term in Eq. (1).
Using supervised online learning of realizable rules and binary classification (Fig. 2b,c), we first examined the impact of the timedependent adaptation steps (Eq. 1) on accelerating biological learning processes. The teacher provided the student asynchronousinput and binaryoutput relations^{8}, where both had the same architecture of the simplest classifier, the perceptron^{9}, and the output nodes were represented by a leaky integrateandfire neuron^{10}. Two scenarios were examined: synaptic adaptation and dendritic adaptation (Fig. 2b,c and Materials and Methods). Results clearly indicate that the generalization error, ε_{g}, of the experimentallyinspired timedependent η (Eq. 1) substantially outperformed the fixed η scenario (Fig. 2b,c). This accelerated learning stems from the fact that weights in synaptic learning converge to the extreme limits, vanishing or abovethreshold weights^{7}. Hence, the learning stepsizes in coherent dynamics increased toward these extremes, accelerating the learning scenario (Fig. 2b). Similarly, in the dendritic case (Fig. 2c), weights oscillated and synchronized via repeatedly hitting the boundary values^{7}. Hence, a faster decay of ε_{g} also resulted from learning step acceleration (Fig. 2c).
Next, we examined the experimentallyinspired timedependent learning step mechanism on the supervised learning of an unrealizable rule using the MNIST database^{11} tested on a neural network. This database consists of a large number of examples of handwritten digits (Fig. 3a) and is commonly used as a prototypical problem for quantifying the generalization performance of machine learning algorithms for various image processing tasks. In this study we use a small subset of the MNIST database without any data extension methods^{12,13,14}. The commonly used trained networks consisted of 784 inputs representing the 28 × 28 pixels of a digit, one hidden layer (30 units in this study), and ten outputs representing the labels (Fig. 3a). The commonly used learning approach is the backpropagation strategy^{15}:
where weight at timestep \(t\), \({W}^{t}\), is modified with a stepsize η towards the minus sign of the gradient of the cost function, C. An improved approach is the momentum strategy^{5,16,17} and regularization of the weights^{18,19}:
where the momentum, μ, and the regularization, α, are constants in the region [0, 1] and \({\eta }_{0}\) is a constant. We optimized the performance of the momentum strategy (Eq. 3) over \((\mu ,\,\alpha ,\,{\eta }_{0})\) for a limited training dataset using the crossentropy cost function (Materials and Methods) and compared its performance with the following two experimentallyinspired learning mechanisms consisting of timedependent η.
In the first approach, acceleration, the timedependent η, and the update rules for weight are given by these equations:
where τ is the positive decaying factor, \({A}_{1}\) and \({\beta }_{1}\) are constants representing the amplitude and the gain between the input and the hidden layers, respectively, and \({A}_{2}\) and \({\beta }_{2}\) represent the same between the hidden and the output layers. It is evident that coherent consecutive gradients of weight, i.e., with the same sign, increased its conjugate η. Note, in the limit \(\beta \to \infty \), the equation for η was simplified, \({\eta }^{t+1}={\eta }^{t}\cdot \exp (\tau )+A\cdot sign({{\nabla }}_{{{\rm{W}}}^{{\rm{t}}}}C)\). The second approach, advanced acceleration, combines the two previous approaches (Eqs. 3, 4):
For the two experimentallyinspired acceleratedapproaches (Eqs. 4, 5), the changes in weights and \(\eta \) depend on the higher moments of gradients, in contrast with the linear dependence of the momentum approach (Eq. 3). Given a limited subset of the dataset examples the performance of the accelerated approaches was maximized over six (Eq. 4) and seven (Eq. 5) parameters (Materials and Methods).
The online training set consisted of 300 randomly chosen examples: each label appeared 30 times within a random order. After 300 learning steps, the accelerated approaches outperformed the momentum method by more than 25%, and the test accuracy increased from about 0.43 to 0.54, respectively (Fig. 3b). Note, the acceleration approach (Eq. 4) showed similar performance to the advanced acceleration approach (Eq. 5). These improved results were found to be robust also for online training based on 6000 examples (30 batches of size 200) (Fig. 3c) and 1200 examples (24 batches of size 50) (Fig. 3d). For both cases (Fig. 3c,d), the advanced acceleration approach (Eq. 5) offered the best performance. We repeated the training using the same 60 examples 5 times (60 × 5 = 300 training in total); compared with using 300 examples for training once, the 60 × 5 approach yielded better performance: using the advanced acceleration approach test accuracy increased from 0.54 to 0.57 (Fig. 3e and Supplementary Fig. 2). For a given number of network updates, results demonstrate that smaller example sets yield more information. This result stems from the random training order of a randomly selected small dataset, e.g., 60 or 300 examples, consisting of a balanced appearance for each label. Around equalized trained label appearances, there are more temporal fluctuations for a dataset involving 300 examples with 30 appearances for each label than for one involving 5 sets of 60 examples with 6 appearances for each label. Indeed, for a training set of 300 distinct examples composed of 5 subsets of 60 balanced examples, a test accuracy of 0.57 was achieved (Supplementary Fig. 3), and for one with 30 subsets of 10 examples where each label appears once, the test accuracy increased further to 0.67 and to 0.7 for a fixed label order (Supplementary Fig. 4). Results indicate that in order to maximize the test accuracy for online scenarios and especially for small datasets, the balanced set of examples and their balanced temporal training order are important ingredients.
Conclusions
Based on increased η with coherent consecutive gradients, the braininspired acceleratedlearning mechanism outperforms existing common ML strategies for small sets of training examples^{20}. Consistent results occur across various cost functions, e.g., square costfunction, however, with a relatively diminished performance (Fig. 3f). Because the performance maximization for a given dataset depends on the selected acceleration approach (Fig. 3b,c), adapting the learning approach during the training process may improve performance. Nevertheless, in addition to possible advanced nonlinear functions for updating η, given the number of network updates, the ultimate scheduling of acceleration approaches and the ordering of trained examples to maximize the performance deserves further research. The presented bridge from experimental neuroscience to ML is expected to further advance decision making using limited databases, which is the reality in many aspects of human activity^{21}, robotic control^{22,23}, and network optimization^{24,25}.
Materials and Methods
InVitro experiments
The experimental methods are similar to our previous studies^{6,7} and only the modifications are presented. All procedures were in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and BarIlan University Guidelines for the Use and Care of Laboratory Animals in Research and were approved and supervised by the BarIlan University Animal Care and Use Committee.
Experiments protocol
The details of the experimental protocol of Figure 1 are as follows. The neuronal response latency (NRL) was used to accurately adjust the timelag between intracellular evoked spikes or EPSPs originated from consecutive intra and extra cellular stimulations to be in the range of 1–4 ms. We note that an abovethreshold extracellular stimulation given shortly, e.g. 2 ms, after an abovethreshold intracellular stimulation, does not result in an evoked spike, and can be used to enhance adaptation. The thresholds and NRL were rechecked at the end of the experiment, in order to ensure their stability.
Statistical analysis
The demonstrated results were repeated tens of times on many cultures.
Simulations of biological neural networks
The simulation methods are similar to our previous studies^{8} and only the used parameters and modifications are presented.
Inputs generation
Each input was composed of N/2 randomly stimulated input units. For each stimulated unit a random delay and a stimulation amplitude were chosen from given distributions. The delays were randomly chosen from a uniform distribution with a resolution of 1 (0.001) ms, such that the average timelag between two consecutive stimulations was 2 (10) ms for the synaptic (dendritic) scenario. Stimulation amplitudes were randomly chosen from a uniform distribution in the range [0.8, 1.2]. Note that the reported results are qualitatively robust to the scenario where all the nonzero amplitudes equal 1. In the dendritic scenario, the five \({W}_{m}\) connected to the same dendrite were stimulated sequentially in a random order and with an average timelag of 10 ms between consecutive stimulations.
Calculating the generalization error
The estimation consisted of up to 20,000 inputs presented to the teacher and the student, where each input generates about 30/200 evoked spikes for the synaptic/dendritic scenario. The generalization error is defined as
The details of the simulations in Figure 2 are as follows. Panel B: \(\{{W}_{m}\}\) were chosen from a uniform distribution in the range [0.1, 0.2]. The adaptation and learning steps were A = 0.05 and \({\rm{\eta }}\) = 1/1000, respectively. \({W}_{m}\) was bounded from above by 1.5 and from below by 10^{−4}. The fixed learning rate was compared to the accelerating method using adaptive learning step:
using \(\tau \) = 0.1, B = 0.01 and \(\eta \) was initiated as 1/1000.
Panel C: \(\{{W}_{m}\}\) were chosen from a uniform distribution in the range [0.1, 0.9] and then were normalized to a mean equals to 0.5. \(\{{J}_{i}\}\) were chosen from a uniform distribution in the range [0.5, 1.5]. Stimulations with low amplitudes (0.01) were given to the N/2 unstimulated input units, resulting in nonfrozen \({J}_{i}\). The adaptation and learning steps were A = 0.05 and \({\rm{\eta }}\) = 1/1000, respectively. \({J}_{i}\) was bounded from below by 0.1 and from above by 3. The fixed learning rate was compared to the accelerating method using adaptive learning step:
using \(\tau \) = 0.1, B = 0.01 and \(\eta \) was initiated as 1/1000.
Simulations of neural network
Architecture
The feedforward neural network contains 784 input units, 30 hidden units and 10 output units in a fully connected architecture. Each unit in the hidden and the output layers has an additional input from a bias unit. Weights from the input layer to the hidden layer, W_{1}, and from the hidden layer to the output layer, W_{2}, were randomly chosen from a Gaussian distribution with a zero average and standard deviation equals 1. All weights were normalized such that all input weights to each hidden unit have an average equals 0 and a STD equals 1. The initial value of the bias of each weights was set to 1. We trained the network on the handwritten digits dataset, MNIST, using gradient descent. The inputs, examples from the train dataset, contain 784 pixel values in the range [0, 255]. We normalized the inputs such that the average and the STD are equal to 0 and 1, respectively.
Forward propagation
The output of a single unit in the hidden layer, \({a}_{j}^{1}\), was calculated as:
where \({W}_{ij}^{1}\) is the weight from the i^{th} input to the j^{th} hidden unit, \({X}_{i}\) is the i^{th} input, and b^{1}_{j} is the bias for the j^{th} hidden unit.
For the output layer, the output of a single unit, \({a}_{j}^{2}\) was calculated as:
where \({W}_{ij}^{2}\) is the weight from the i^{th} hidden unit to the j^{th} output unit, \({a}_{i}^{1}\) is the output of the i^{th} unit in the hidden layer, and b^{2}_{j} is the bias for the j^{th} output unit.
Back propagation
We used two different cost functions; the first was the cross entropy:
and the second was the mean square error (MSE):
where y are the desired labels and \(a\) stands for the current 10 output units of the output layer. The summation is over all training examples, N.
The backpropagation method computes the gradient for each weight with respect to the chosen cost function. The weights and biases were updated according to 3 different methods:
 (1)
Momentum
The weights update:
$${W}^{t+1}=(1\alpha )\cdot {W}^{t}+{V}^{t+1}$$$${V}^{t+1}=\mu \cdot {V}^{t}{\eta }_{0}\cdot {\nabla }_{{{\rm{W}}}^{{\rm{t}}}}C$$where t is the discrete timestep W are the weights, α is a regularization constant, \(\eta \) is the fixed learning rate, and \({\nabla }_{{{\rm{W}}}^{{\rm{t}}}}C\) is the gradient of the cost function for each weight at time t. V was initialized as: \({\eta }_{0}\cdot {\nabla }_{{\rm{W}}}{C}_{first}\), where \({\nabla }_{W}{C}_{first}\) is the first computed gradient and the biases update:
$${V}_{b}^{t+1}=\mu \cdot {V}_{b}^{t}{\eta }_{0}\cdot {\nabla }_{{b}^{t}}C$$$${b}^{t+1}={b}^{t}+{V}_{b}^{t+1}$$where \({\nabla }_{b}C\) is the gradient of the cost function of each bias with respect to its weight, b, and V_{b} was initialized as: \(\eta \cdot {\nabla }_{{\rm{b}}}{C}_{first}\), where \({\nabla }_{b}{C}_{first}\) is the first computed bias gradient.
 (2)
Acceleration
The weights update:
$${W}^{t+1}=(1\alpha )\cdot {W}^{t}{\eta }^{t+1}\cdot {\nabla }_{{W}^{t}}C$$$${\eta }^{t+1}={\eta }^{t}\cdot {e}^{\tau }+{A}_{1/2}\cdot tanh({\beta }_{1/2}\cdot {\nabla }_{{W}^{t}}C)$$where \(\eta \) is defined for each weight, \(\,{A}_{1}\) and \({\beta }_{1}\) are constants representing the amplitude and the gain between the input and the hidden layers, respectively, and \({A}_{2}\) and \({\beta }_{2}\) represent the same between the hidden and the output layers. \(\eta \) was initialized as: \({A}_{1/2}\cdot tanh({\beta }_{1/2}\cdot {\nabla }_{W}{C}_{first})\), where \({\nabla }_{W}{C}_{first}\) is the first computed gradient and the biases update:
$${b}^{t+1}={b}^{t}{\eta }_{b}^{t+1}\cdot {\nabla }_{{b}^{t}}C$$$${\eta }_{b}^{t+1}={\eta }_{b}^{t}\cdot {e}^{\tau }+{A}_{1/2}\cdot tanh({\beta }_{1/2}\cdot {\nabla }_{{b}^{t}}C)$$  (3)
Advanced acceleration:
The weights update
and the biases update:
Testing the network
The network classification accuracy was tested on the MNIST dataset for testing, containing 10,000 inputs. The test inputs were also normalized to have an average of each equals to 0 and a STD equals to 1.
Optimization
For each update method the parameters were chosen to maximize the test accuracy. For optimization we first used a grid of the adjustable parameters followed by a fine tuning with higher resolution for each parameter. The optimization was performed over 3 parameters for the momentum method (\(\mu ,{\eta }_{0},\alpha \)), 6 parameters for the acceleration method (\({A}_{1},{A}_{2},{\beta }_{1},{\beta }_{2},\tau ,\,\alpha \)) and for 7 parameters for the advanced acceleration method (\({A}_{1},{A}_{2},{\beta }_{1},{\beta }_{2},\tau ,\,\alpha ,\mu \)).
The details of the simulations in Figure 3 are as follows. Panel B: The feed forward neural network was presented with 300 examples with equally number of appearance of each digit. The cross entropy cost function was used and the following parameters for each method: momentum with \(\mu =0.9,{\eta }_{0}=0.02,\alpha =0\), acceleration with \({A}_{1}=0.1,{A}_{2}=0.27,{\beta }_{1}=\,5000,{\beta }_{2}=\,900,\tau =\,0.598,\,\alpha =0.003\), and advanced acceleration with \({A}_{1}=0.07,{A}_{2}=0.07,{\beta }_{1}=\infty ,{\beta }_{2}=\,\infty ,\tau =0.29,\,\alpha =\,0.005,\mu =0.5\). Results are presented as the average of 100 different runs, and typical error bars are presented for the last point.
Panel C: The feed forward neural network was presented with 6000 examples randomly taken from the 60000 examples in the training dataset, and presented to the network with 30 minibatches of size 200. The cross entropy cost function was used and the following parameters for each method: momentum with \(\mu =0.68,{\eta }_{0}=0.65,\alpha =0.08\), acceleration with \({A}_{1}=3,{A}_{2}=0.7,{\beta }_{1}=\,1500,{\beta }_{2}=\,1000,\tau =\,0.4,\)\(\alpha =0.065\), and advanced acceleration with \({A}_{1}=0.5,{A}_{2}=0.5,{\beta }_{1}=\infty ,{\beta }_{2}=\,\infty ,\tau =0.1,\,\alpha =\,0.1,\)\(\mu =0.55\). Results are presented as the average of 100 different runs, and typical error bars are presented for the last point.
Panel D: The feed forward neural network was presented with 1200 examples randomly taken from the 60000 examples in the training dataset, and presented to the network with 24 minibatches of size 50. The cross entropy cost function was used and the following parameters for each method: momentum with \(\mu =0.75,\)\({\eta }_{0}=0.6,\alpha =0.11\), acceleration with \({A}_{1}=1.15,{A}_{2}=0.6,{\beta }_{1}=\,4500,{\beta }_{2}=\,3500,\tau =\,0.1,\,\alpha =0.055\), and advanced acceleration with \({A}_{1}=0.55,{A}_{2}=0.55,{\beta }_{1}=\infty ,{\beta }_{2}=\,\infty ,\tau =0.1,\,\alpha =\,0.1,\mu =0.55\). Results are presented as the average of 100 different runs, and typical error bars are presented for the last point.
Panel E: The feed forward neural network was presented with 60 with equally number of appearance of each digit. The 60 examples were presented to the network 5 times. The cross entropy cost function was used and the following parameters for each method: momentum with \(\mu =0.87,{\eta }_{0}=0.035,\alpha =0.005\), acceleration with \({A}_{1}=0.11,{A}_{2}=0.26,{\beta }_{1}=\,2000,{\beta }_{2}=\,2500,\tau =\,0.45,\,\alpha =0.008\), and advanced acceleration with \({A}_{1}=0.035,{A}_{2}=0.02,{\beta }_{1}=4500,{\beta }_{2}=\,5,\tau =0.0619,\alpha =\,0.01,\mu =0.6\). Results are presented as the average of 100 different runs, and typical error bars are presented for the last point.
Panel F: The feed forward neural network was presented with 300 examples with equally number of appearance of each digit. The meansquareerror cost function was used and the following parameters for each method: momentum with \(\mu =0.6,{\eta }_{0}=0.35,\alpha =0.005\), acceleration with \({A}_{1}=0.95,{A}_{2}=0.25,{\beta }_{1}=\,5000,\)\({\beta }_{2}=\,40,\tau =\,0.15,\alpha =0.008\), and advanced acceleration with \({A}_{1}=0.06,{A}_{2}=0.09,{\beta }_{1}=2100,{\beta }_{2}=\,1,\)\(\tau =0.015,\,\alpha =\,0.005,\mu =0.8\). Results are presented as the average of 100 different runs, and typical error bars are presented for the last point.
The feed forward neural network was presented with 300 examples with equally number of appearance of each digit. The examples are composed of 5 subsets of 60 balanced examples, where in every subset each label appears exactly 6 times. The cross entropy cost function was used and the following parameters for each method: momentum with \(\mu =0.87,{\eta }_{0}=0.035,\alpha =0.005\), acceleration with \({A}_{1}=0.11,{A}_{2}=0.26,{\beta }_{1}=\,2000,{\beta }_{2}=\,2500,\tau =\,0.45,\,\alpha =0.008\), and advanced acceleration with \({A}_{1}=0.035,{A}_{2}=0.02,{\beta }_{1}=4500,{\beta }_{2}=\,5,\tau =0.0619,\,\alpha =\,0.01,\mu =0.6\). Results are presented as the average of 100 different runs, and typical error bars are presented for the last point.
The feedforward neural network was presented with 300 examples with equally number of appearance of each digit. The examples are composed of 30 subsets of 10 balanced examples, where in every subset each label appears exactly one time. Results are also presented for the advanced acceleration method where for each subset of 10 examples, the labels were in a fixed order. The cross entropy cost function was used and the following parameters for each method: momentum with \(\mu =0.87,{\eta }_{0}=0.035,\alpha =0.005\), acceleration with \({A}_{1}=0.11,{A}_{2}=0.26,{\beta }_{1}=\,2000,{\beta }_{2}=\,2500,\tau =\,0.45,\,\alpha =0.008\), and advanced acceleration with \({A}_{1}=0.035,{A}_{2}=0.02,{\beta }_{1}=4500,{\beta }_{2}=\,5,\tau =0.0619,\alpha =\,0.01,\mu =0.6\). Results are presented as the average of 100 different runs, and typical error bars are presented for the last point.
Change history
04 June 2020
An amendment to this paper has been published and can be accessed via a link at the top of the paper.
References
Hebb, D. The organization of behavior. A Neuropsychological (1949).
Dan, Y. & Poo, M.m Hebbian depression of isolated neuromuscular synapses in vitro. Science 256, 1570–1573 (1992).
Turrigiano, G. G. & Nelson, S. B. Homeostatic plasticity in the developing nervous system. Nature Reviews Neuroscience 5, 97 (2004).
Markram, H., Gerstner, W. & Sjöström, P. J. Spiketimingdependent plasticity: a comprehensive overview. Frontiers in synaptic neuroscience 4, 2 (2012).
Perantonis, S. J. & Karras, D. A. An efficient constrained learning algorithm with momentum acceleration. Neural Networks 8, 237–249 (1995).
Sardi, S., Vardi, R., Sheinin, A., Goldental, A. & Kanter, I. New Types of Experiments Reveal that a Neuron Functions as Multiple Independent Threshold Units. Sci RepUk 7, 18036 (2017).
Sardi, S. et al. Adaptive nodes enrich nonlinear cooperative learning beyond traditional adaptation by links. Sci RepUk 8, 5100 (2018).
Uzan, H., Sardi, S., Goldental, A., Vardi, R. & Kanter, I. Biological learning curves outperform existing ones in artificial intelligence algorithms. Sci RepUk 9, 11558, https://doi.org/10.1038/s41598019480164 (2019).
Watkin, T. L., Rau, A. & Biehl, M. The statistical mechanics of learning a rule. Reviews of Modern Physics 65, 499 (1993).
Brette, R. & Gerstner, W. Adaptive exponential integrateandfire model as an effective description of neuronal activity. Journal of neurophysiology 94, 3637–3642 (2005).
LeCun, Y. et al. Learning algorithms for classification: A comparison on handwritten digit recognition. Neural networks: the statistical mechanics perspective 261, 276 (1995).
Zhang, Y. & Ling, C. A strategy to apply machine learning to small datasets in materials science. Npj Computational Materials 4, 25 (2018).
Shorten, C. & Khoshgoftaar, T. M. A survey on image data augmentation for deep learning. Journal of Big Data 6, 60 (2019).
Yosinski, J., Clune, J., Bengio, Y. & Lipson, H. In Advances in neural information processing systems. 3320–3328 (2014).
LeCun, Y., Bengio, Y. & Hinton, G. Deep learning. Nature 521, 436 (2015).
Sutskever, I., Martens, J., Dahl, G. & Hinton, G. In International conference on machine learning. 1139–1147 (2013).
LeCun, Y. A., Bottou, L., Orr, G. B. & Müller, K.R. In Neural networks: Tricks of the trade 9–48 (Springer, 2012).
Jin, Y., Okabe, T. & Sendhoff, B. In Proceedings of the 2004 Congress on Evolutionary Computation (IEEE Cat. No. 04TH8753). 18 (IEEE).
Bishop, C. M. Neural networks for pattern recognition. (Oxford university press, 1995).
Hoffmann, J. et al. Machine learning in a datalimited regime: Augmenting experiments with synthetic data uncovers order in crumpled sheets. Science advances 5, eaau6792 (2019).
Kearns, M., Suri, S. & Montfort, N. An experimental study of the coloring problem on human subject networks. Science 313, 824–827 (2006).
Edelman, B. et al. Noninvasive neuroimaging enhances continuous neural tracking for robotic device control. Science Robotics 4, eaaw6844 (2019).
Saridis, G. Intelligent robotic control. IEEE Transactions on Automatic Control 28, 547–557 (1983).
Mateo, D., Horsevad, N., Hassani, V., Chamanbaz, M. & Bouffanais, R. Optimal network topology for responsive collective behavior. Science Advances 5, eaau0999 (2019).
Liu, L., Cheng, Y., Cai, L., Zhou, S. & Niu, Z. In 2017 IEEE international conference on communications (ICC). 16 (IEEE).
Acknowledgements
We would like to thank Editage (www.editage.com) for English language editing.
Author information
Authors and Affiliations
Contributions
S.S. and R.V. contributed equally to all experimental result aspects while Y.T. prepared the tissue cultures and materials for the experiments. S.S. and Y.M. contributed equally to the simulation results. S.H. confirmed some of the simulation results. A.G. contributed to conceptualization and for the data analysis. I.K. initiated the study and supervised all aspects of the work. All authors discussed the results and commented on the manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Sardi, S., Vardi, R., Meir, Y. et al. Brain experiments imply adaptation mechanisms which outperform common AI learning algorithms. Sci Rep 10, 6923 (2020). https://doi.org/10.1038/s41598020637555
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598020637555
This article is cited by

Enhancing the accuracies by performing pooling decisions adjacent to the output layer
Scientific Reports (2023)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.