Limits...
Deep Neural Networks with Multistate Activation Functions.

Cai C, Xu Y, Ke D, Su K - Comput Intell Neurosci (2015)

Bottom Line: Experimental results on the TIMIT corpus reveal that, on speech recognition tasks, DNNs with MSAFs perform better than the conventional DNNs, getting a relative improvement of 5.60% on phoneme error rates.Further experiments also reveal that mean-normalised SGD facilitates the training processes of DNNs with MSAFs, especially when being with large training sets.The models can also be directly trained without pretraining when the training set is sufficiently large, which results in a considerable relative improvement of 5.82% on word error rates.

View Article: PubMed Central - PubMed

Affiliation: School of Technology, Beijing Forestry University, No. 35 Qinghuadong Road, Haidian District, Beijing 100083, China.

ABSTRACT
We propose multistate activation functions (MSAFs) for deep neural networks (DNNs). These MSAFs are new kinds of activation functions which are capable of representing more than two states, including the N-order MSAFs and the symmetrical MSAF. DNNs with these MSAFs can be trained via conventional Stochastic Gradient Descent (SGD) as well as mean-normalised SGD. We also discuss how these MSAFs perform when used to resolve classification problems. Experimental results on the TIMIT corpus reveal that, on speech recognition tasks, DNNs with MSAFs perform better than the conventional DNNs, getting a relative improvement of 5.60% on phoneme error rates. Further experiments also reveal that mean-normalised SGD facilitates the training processes of DNNs with MSAFs, especially when being with large training sets. The models can also be directly trained without pretraining when the training set is sufficiently large, which results in a considerable relative improvement of 5.82% on word error rates.

No MeSH data available.


An example of classification. The primary task is to classify these 4 patterns and mark them, respectively.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4581500&req=5

fig5: An example of classification. The primary task is to classify these 4 patterns and mark them, respectively.

Mentions: Figure 5 illustrates a classification task. There are four different patterns in this figure. Each pattern corresponds to two values, which are its horizontal and longitudinal coordinates, respectively. According to their coordinates, they will be divided into four different classes. If we use a neural network and the logistic function to classify them, the neural network will have a considerable number of units, especially in its output layer, due to the fact that the logistic function only has two states.


Deep Neural Networks with Multistate Activation Functions.

Cai C, Xu Y, Ke D, Su K - Comput Intell Neurosci (2015)

An example of classification. The primary task is to classify these 4 patterns and mark them, respectively.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4581500&req=5

fig5: An example of classification. The primary task is to classify these 4 patterns and mark them, respectively.
Mentions: Figure 5 illustrates a classification task. There are four different patterns in this figure. Each pattern corresponds to two values, which are its horizontal and longitudinal coordinates, respectively. According to their coordinates, they will be divided into four different classes. If we use a neural network and the logistic function to classify them, the neural network will have a considerable number of units, especially in its output layer, due to the fact that the logistic function only has two states.

Bottom Line: Experimental results on the TIMIT corpus reveal that, on speech recognition tasks, DNNs with MSAFs perform better than the conventional DNNs, getting a relative improvement of 5.60% on phoneme error rates.Further experiments also reveal that mean-normalised SGD facilitates the training processes of DNNs with MSAFs, especially when being with large training sets.The models can also be directly trained without pretraining when the training set is sufficiently large, which results in a considerable relative improvement of 5.82% on word error rates.

View Article: PubMed Central - PubMed

Affiliation: School of Technology, Beijing Forestry University, No. 35 Qinghuadong Road, Haidian District, Beijing 100083, China.

ABSTRACT
We propose multistate activation functions (MSAFs) for deep neural networks (DNNs). These MSAFs are new kinds of activation functions which are capable of representing more than two states, including the N-order MSAFs and the symmetrical MSAF. DNNs with these MSAFs can be trained via conventional Stochastic Gradient Descent (SGD) as well as mean-normalised SGD. We also discuss how these MSAFs perform when used to resolve classification problems. Experimental results on the TIMIT corpus reveal that, on speech recognition tasks, DNNs with MSAFs perform better than the conventional DNNs, getting a relative improvement of 5.60% on phoneme error rates. Further experiments also reveal that mean-normalised SGD facilitates the training processes of DNNs with MSAFs, especially when being with large training sets. The models can also be directly trained without pretraining when the training set is sufficiently large, which results in a considerable relative improvement of 5.82% on word error rates.

No MeSH data available.