Limits...
Multiclass Classification by Adaptive Network of Dendritic Neurons with Binary Synapses Using Structural Plasticity.

Hussain S, Basu A - Front Neurosci (2016)

Bottom Line: The performance of the model is evaluated on classification of handwritten digits from the benchmark MNIST dataset and compared with other spike classifiers.For VLSI implementations, we show that the reduced synaptic memory can save upto 4X area compared to conventional crossbar topologies.Finally, we also present a biologically realistic spike-based version for calculating the correlations required by the structural learning rule and demonstrate the correspondence between the rate-based and spike-based methods of learning.

View Article: PubMed Central - PubMed

Affiliation: School of Electrical and Electronic Engineering, Nanyang Technological University Singapore, Singapore.

ABSTRACT
The development of power-efficient neuromorphic devices presents the challenge of designing spike pattern classification algorithms which can be implemented on low-precision hardware and can also achieve state-of-the-art performance. In our pursuit of meeting this challenge, we present a pattern classification model which uses a sparse connection matrix and exploits the mechanism of nonlinear dendritic processing to achieve high classification accuracy. A rate-based structural learning rule for multiclass classification is proposed which modifies a connectivity matrix of binary synaptic connections by choosing the best "k" out of "d" inputs to make connections on every dendritic branch (k < < d). Because learning only modifies connectivity, the model is well suited for implementation in neuromorphic systems using address-event representation (AER). We develop an ensemble method which combines several dendritic classifiers to achieve enhanced generalization over individual classifiers. We have two major findings: (1) Our results demonstrate that an ensemble created with classifiers comprising moderate number of dendrites performs better than both ensembles of perceptrons and of complex dendritic trees. (2) In order to determine the moderate number of dendrites required for a specific classification problem, a two-step solution is proposed. First, an adaptive approach is proposed which scales the relative size of the dendritic trees of neurons for each class. It works by progressively adding dendrites with fixed number of synapses to the network, thereby allocating synaptic resources as per the complexity of the given problem. As a second step, theoretical capacity calculations are used to convert each neuronal dendritic tree to its optimal topology where dendrites of each class are assigned different number of synapses. The performance of the model is evaluated on classification of handwritten digits from the benchmark MNIST dataset and compared with other spike classifiers. We show that our system can achieve classification accuracy within 1 - 2% of other reported spike-based classifiers while using much less synaptic resources (only 7%) compared to that used by other methods. Further, an ensemble classifier created with adaptively learned sizes can attain accuracy of 96.4% which is at par with the best reported performance of spike-based classifiers. Moreover, the proposed method achieves this by using about 20% of the synapses used by other spike algorithms. We also present results of applying our algorithm to classify the MNIST-DVS dataset collected from a real spike-based image sensor and show results comparable to the best reported ones (88.1% accuracy). For VLSI implementations, we show that the reduced synaptic memory can save upto 4X area compared to conventional crossbar topologies. Finally, we also present a biologically realistic spike-based version for calculating the correlations required by the structural learning rule and demonstrate the correspondence between the rate-based and spike-based methods of learning.

No MeSH data available.


Related in: MedlinePlus

(A) The number of dendrites learned using adaptive scheme-1 (top) and the resulting confusion matrix (bottom) showing higher accuracy for easier classes (0, 1, 6) and lower accuracy for difficult digits (8, 9). (B) Number of dendrites learned using adaptive scheme-2 (top) demonstrating all the class neurons adding dendrites when required. Confusion matrix (bottom) showing improved accuracy for all the classes. Training set consists of 20, 000 binary digit samples and testing done on 10, 000 jittered single spikes (Δ = 10 ms). Model was initialized with m = 5 dendrites for all the classes.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4814530&req=5

Figure 6: (A) The number of dendrites learned using adaptive scheme-1 (top) and the resulting confusion matrix (bottom) showing higher accuracy for easier classes (0, 1, 6) and lower accuracy for difficult digits (8, 9). (B) Number of dendrites learned using adaptive scheme-2 (top) demonstrating all the class neurons adding dendrites when required. Confusion matrix (bottom) showing improved accuracy for all the classes. Training set consists of 20, 000 binary digit samples and testing done on 10, 000 jittered single spikes (Δ = 10 ms). Model was initialized with m = 5 dendrites for all the classes.

Mentions: We have used the adaptive learning scheme in which P = 20, 000 patterns were used for training by initializing the network with m = 5 dendrites in each dendritic tree of a class, and then adaptively increasing m in a class-specific manner. The learning process was stopped when 150 minima were encountered and therefore, each adaptively learned classifier took different number of iterations to complete 150 minima. If dendrites are added to only 5 worst-performing classes whenever their learning slows down, referred to as scheme-1, the accuracy obtained on 10, 000 test single spike inputs is 92.1%. The number of dendrites learned by each class in one simulation run is shown by the bar plot at the top of Figure 6A. It can be seen that the digits “2,” “3,” “5,” “8,” and “9” use most of the dendrites and hence are most difficult to learn while “0,” “1,” and “6” are the easier ones requiring only a small number of dendrites. The confusion matrix at the bottom of Figure 6A shows the classification accuracy when an actual digit (column-wise) is predicted by the model as represented row-wise. We can see that the neurons for easier digits “0,” “1,” and “6” can attain good accuracy by utilizing small number of dendrites whereas the difficult digits like “8” and “9” exhibit lower accuracy, where “9” is mostly misclassified as “4” and “7” having similar features.


Multiclass Classification by Adaptive Network of Dendritic Neurons with Binary Synapses Using Structural Plasticity.

Hussain S, Basu A - Front Neurosci (2016)

(A) The number of dendrites learned using adaptive scheme-1 (top) and the resulting confusion matrix (bottom) showing higher accuracy for easier classes (0, 1, 6) and lower accuracy for difficult digits (8, 9). (B) Number of dendrites learned using adaptive scheme-2 (top) demonstrating all the class neurons adding dendrites when required. Confusion matrix (bottom) showing improved accuracy for all the classes. Training set consists of 20, 000 binary digit samples and testing done on 10, 000 jittered single spikes (Δ = 10 ms). Model was initialized with m = 5 dendrites for all the classes.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4814530&req=5

Figure 6: (A) The number of dendrites learned using adaptive scheme-1 (top) and the resulting confusion matrix (bottom) showing higher accuracy for easier classes (0, 1, 6) and lower accuracy for difficult digits (8, 9). (B) Number of dendrites learned using adaptive scheme-2 (top) demonstrating all the class neurons adding dendrites when required. Confusion matrix (bottom) showing improved accuracy for all the classes. Training set consists of 20, 000 binary digit samples and testing done on 10, 000 jittered single spikes (Δ = 10 ms). Model was initialized with m = 5 dendrites for all the classes.
Mentions: We have used the adaptive learning scheme in which P = 20, 000 patterns were used for training by initializing the network with m = 5 dendrites in each dendritic tree of a class, and then adaptively increasing m in a class-specific manner. The learning process was stopped when 150 minima were encountered and therefore, each adaptively learned classifier took different number of iterations to complete 150 minima. If dendrites are added to only 5 worst-performing classes whenever their learning slows down, referred to as scheme-1, the accuracy obtained on 10, 000 test single spike inputs is 92.1%. The number of dendrites learned by each class in one simulation run is shown by the bar plot at the top of Figure 6A. It can be seen that the digits “2,” “3,” “5,” “8,” and “9” use most of the dendrites and hence are most difficult to learn while “0,” “1,” and “6” are the easier ones requiring only a small number of dendrites. The confusion matrix at the bottom of Figure 6A shows the classification accuracy when an actual digit (column-wise) is predicted by the model as represented row-wise. We can see that the neurons for easier digits “0,” “1,” and “6” can attain good accuracy by utilizing small number of dendrites whereas the difficult digits like “8” and “9” exhibit lower accuracy, where “9” is mostly misclassified as “4” and “7” having similar features.

Bottom Line: The performance of the model is evaluated on classification of handwritten digits from the benchmark MNIST dataset and compared with other spike classifiers.For VLSI implementations, we show that the reduced synaptic memory can save upto 4X area compared to conventional crossbar topologies.Finally, we also present a biologically realistic spike-based version for calculating the correlations required by the structural learning rule and demonstrate the correspondence between the rate-based and spike-based methods of learning.

View Article: PubMed Central - PubMed

Affiliation: School of Electrical and Electronic Engineering, Nanyang Technological University Singapore, Singapore.

ABSTRACT
The development of power-efficient neuromorphic devices presents the challenge of designing spike pattern classification algorithms which can be implemented on low-precision hardware and can also achieve state-of-the-art performance. In our pursuit of meeting this challenge, we present a pattern classification model which uses a sparse connection matrix and exploits the mechanism of nonlinear dendritic processing to achieve high classification accuracy. A rate-based structural learning rule for multiclass classification is proposed which modifies a connectivity matrix of binary synaptic connections by choosing the best "k" out of "d" inputs to make connections on every dendritic branch (k < < d). Because learning only modifies connectivity, the model is well suited for implementation in neuromorphic systems using address-event representation (AER). We develop an ensemble method which combines several dendritic classifiers to achieve enhanced generalization over individual classifiers. We have two major findings: (1) Our results demonstrate that an ensemble created with classifiers comprising moderate number of dendrites performs better than both ensembles of perceptrons and of complex dendritic trees. (2) In order to determine the moderate number of dendrites required for a specific classification problem, a two-step solution is proposed. First, an adaptive approach is proposed which scales the relative size of the dendritic trees of neurons for each class. It works by progressively adding dendrites with fixed number of synapses to the network, thereby allocating synaptic resources as per the complexity of the given problem. As a second step, theoretical capacity calculations are used to convert each neuronal dendritic tree to its optimal topology where dendrites of each class are assigned different number of synapses. The performance of the model is evaluated on classification of handwritten digits from the benchmark MNIST dataset and compared with other spike classifiers. We show that our system can achieve classification accuracy within 1 - 2% of other reported spike-based classifiers while using much less synaptic resources (only 7%) compared to that used by other methods. Further, an ensemble classifier created with adaptively learned sizes can attain accuracy of 96.4% which is at par with the best reported performance of spike-based classifiers. Moreover, the proposed method achieves this by using about 20% of the synapses used by other spike algorithms. We also present results of applying our algorithm to classify the MNIST-DVS dataset collected from a real spike-based image sensor and show results comparable to the best reported ones (88.1% accuracy). For VLSI implementations, we show that the reduced synaptic memory can save upto 4X area compared to conventional crossbar topologies. Finally, we also present a biologically realistic spike-based version for calculating the correlations required by the structural learning rule and demonstrate the correspondence between the rate-based and spike-based methods of learning.

No MeSH data available.


Related in: MedlinePlus