Limits...
Robustness of spiking Deep Belief Networks to noise and reduced bit precision of neuro-inspired hardware platforms.

Stromatias E, Neil D, Pfeiffer M, Galluppi F, Furber SB, Liu SC - Front Neurosci (2015)

Bottom Line: Increasingly large deep learning architectures, such as Deep Belief Networks (DBNs) are the focus of current machine learning research and achieve state-of-the-art results in different domains.The on-going work on design and construction of spike-based hardware platforms offers an alternative for running deep neural networks with significantly lower power consumption, but has to overcome hardware limitations in terms of noise and limited weight precision, as well as noise inherent in the sensor signal.Our results demonstrate that spiking DBNs can tolerate very low levels of hardware bit precision down to almost two bits, and show that their performance can be improved by at least 30% through an adapted training mechanism that takes the bit precision of the target platform into account.

View Article: PubMed Central - PubMed

Affiliation: Advanced Processor Technologies Group, School of Computer Science, University of Manchester Manchester, UK.

ABSTRACT
Increasingly large deep learning architectures, such as Deep Belief Networks (DBNs) are the focus of current machine learning research and achieve state-of-the-art results in different domains. However, both training and execution of large-scale Deep Networks require vast computing resources, leading to high power requirements and communication overheads. The on-going work on design and construction of spike-based hardware platforms offers an alternative for running deep neural networks with significantly lower power consumption, but has to overcome hardware limitations in terms of noise and limited weight precision, as well as noise inherent in the sensor signal. This article investigates how such hardware constraints impact the performance of spiking neural network implementations of DBNs. In particular, the influence of limited bit precision during execution and training, and the impact of silicon mismatch in the synaptic weight parameters of custom hybrid VLSI implementations is studied. Furthermore, the network performance of spiking DBNs is characterized with regard to noise in the spiking input signal. Our results demonstrate that spiking DBNs can tolerate very low levels of hardware bit precision down to almost two bits, and show that their performance can be improved by at least 30% through an adapted training mechanism that takes the bit precision of the target platform into account. Spiking DBNs thus present an important use-case for large-scale hybrid analog-digital or digital neuromorphic platforms such as SpiNNaker, which can execute large but precision-constrained deep networks in real time.

No MeSH data available.


Effect of reduced bit precision on firing rates in the DBN and neuron activations. (A) Mean firing rate of each layer of the network for weights with different bit precisions, using an input rate of 1500 Hz. Lower precision levels, which lead to more weights at zero, cause lower firing rates within the network. (B) Distribution of the mean difference between the activation of a neuron with double precision weights and neurons using weights with different bit precision levels. Shown are distributions over all test samples. The difference, although peaked near zero, increases for higher layers, and shows a trend toward reduced activations.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4496577&req=5

Figure 6: Effect of reduced bit precision on firing rates in the DBN and neuron activations. (A) Mean firing rate of each layer of the network for weights with different bit precisions, using an input rate of 1500 Hz. Lower precision levels, which lead to more weights at zero, cause lower firing rates within the network. (B) Distribution of the mean difference between the activation of a neuron with double precision weights and neurons using weights with different bit precision levels. Shown are distributions over all test samples. The difference, although peaked near zero, increases for higher layers, and shows a trend toward reduced activations.

Mentions: However, even with these similar shaped weight distributions, neurons' output firing rates may become dramatically altered by the subtle coercion of weights to become more similar to each other. For this, refer to Figure 6A which shows that for even high levels of quantization, the mean output spike rate per neuron for each of the three layers remains quite constant down to Q3.3, before a clear drop in the mean firing rate is observed. This trend is seen for all three layers, but is stronger in higher layers.


Robustness of spiking Deep Belief Networks to noise and reduced bit precision of neuro-inspired hardware platforms.

Stromatias E, Neil D, Pfeiffer M, Galluppi F, Furber SB, Liu SC - Front Neurosci (2015)

Effect of reduced bit precision on firing rates in the DBN and neuron activations. (A) Mean firing rate of each layer of the network for weights with different bit precisions, using an input rate of 1500 Hz. Lower precision levels, which lead to more weights at zero, cause lower firing rates within the network. (B) Distribution of the mean difference between the activation of a neuron with double precision weights and neurons using weights with different bit precision levels. Shown are distributions over all test samples. The difference, although peaked near zero, increases for higher layers, and shows a trend toward reduced activations.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4496577&req=5

Figure 6: Effect of reduced bit precision on firing rates in the DBN and neuron activations. (A) Mean firing rate of each layer of the network for weights with different bit precisions, using an input rate of 1500 Hz. Lower precision levels, which lead to more weights at zero, cause lower firing rates within the network. (B) Distribution of the mean difference between the activation of a neuron with double precision weights and neurons using weights with different bit precision levels. Shown are distributions over all test samples. The difference, although peaked near zero, increases for higher layers, and shows a trend toward reduced activations.
Mentions: However, even with these similar shaped weight distributions, neurons' output firing rates may become dramatically altered by the subtle coercion of weights to become more similar to each other. For this, refer to Figure 6A which shows that for even high levels of quantization, the mean output spike rate per neuron for each of the three layers remains quite constant down to Q3.3, before a clear drop in the mean firing rate is observed. This trend is seen for all three layers, but is stronger in higher layers.

Bottom Line: Increasingly large deep learning architectures, such as Deep Belief Networks (DBNs) are the focus of current machine learning research and achieve state-of-the-art results in different domains.The on-going work on design and construction of spike-based hardware platforms offers an alternative for running deep neural networks with significantly lower power consumption, but has to overcome hardware limitations in terms of noise and limited weight precision, as well as noise inherent in the sensor signal.Our results demonstrate that spiking DBNs can tolerate very low levels of hardware bit precision down to almost two bits, and show that their performance can be improved by at least 30% through an adapted training mechanism that takes the bit precision of the target platform into account.

View Article: PubMed Central - PubMed

Affiliation: Advanced Processor Technologies Group, School of Computer Science, University of Manchester Manchester, UK.

ABSTRACT
Increasingly large deep learning architectures, such as Deep Belief Networks (DBNs) are the focus of current machine learning research and achieve state-of-the-art results in different domains. However, both training and execution of large-scale Deep Networks require vast computing resources, leading to high power requirements and communication overheads. The on-going work on design and construction of spike-based hardware platforms offers an alternative for running deep neural networks with significantly lower power consumption, but has to overcome hardware limitations in terms of noise and limited weight precision, as well as noise inherent in the sensor signal. This article investigates how such hardware constraints impact the performance of spiking neural network implementations of DBNs. In particular, the influence of limited bit precision during execution and training, and the impact of silicon mismatch in the synaptic weight parameters of custom hybrid VLSI implementations is studied. Furthermore, the network performance of spiking DBNs is characterized with regard to noise in the spiking input signal. Our results demonstrate that spiking DBNs can tolerate very low levels of hardware bit precision down to almost two bits, and show that their performance can be improved by at least 30% through an adapted training mechanism that takes the bit precision of the target platform into account. Spiking DBNs thus present an important use-case for large-scale hybrid analog-digital or digital neuromorphic platforms such as SpiNNaker, which can execute large but precision-constrained deep networks in real time.

No MeSH data available.