ABSTRACT
Learning is a fundamental component of creating intelligent machines. Biological intelligence orchestrates synaptic and neuronal learning at multiple time scales to self-organize populations of neurons for solving complex tasks. Inspired by this, we design and experimentally demonstrate an adaptive hardware architecture Memristive Self-organizing Spiking Recurrent Neural Network (MEMSORN). MEMSORN incorporates resistive memory (RRAM) in its synapses and neurons which configure their state based on Hebbian and Homeostatic plasticity respectively. For the first time, we derive these plasticity rules directly from the statistical measurements of our fabricated RRAM-based neurons and synapses. These "technologically plausible" learning rules exploit the intrinsic variability of the devices and improve the accuracy of the network on a sequence learning task by 30%. Finally, we compare the performance of MEMSORN to a fully-randomly-set-up spiking recurrent network on the same task, showing that self-organization improves the accuracy by more than 15%. This work demonstrates the importance of the device-circuit-algorithm co-design approach for implementing brain-inspired computing hardware.
Subject(s)
Neural Networks, Computer , Synapses , Algorithms , Learning/physiology , Neurons/physiology , Synapses/physiologyABSTRACT
Spike timing-dependent plasticity (STDP), which is widely studied as a fundamental synaptic update rule for neuromorphic hardware, requires precise control of continuous weights. From the viewpoint of hardware implementation, a simplified update rule is desirable. Although simplified STDP with stochastic binary synapses was proposed previously, we find that it leads to degradation of memory maintenance during learning, which is unfavourable for unsupervised online learning. In this work, we propose a stochastic binary synaptic model where the cumulative probability of the weight change evolves in a sigmoidal fashion with potentiation or depression trials, which can be implemented using a pair of switching devices consisting of serially connected multiple binary memristors. As a benchmark test we perform simulations of unsupervised learning of MNIST images with a two-layer network and show that simplified STDP in combination with this model can outperform conventional rules with continuous weights not only in memory maintenance but also in recognition accuracy. Our method achieves 97.3% in recognition accuracy, which is higher than that reported with standard STDP in the same framework. We also show that the high performance of our learning rule is robust against device-to-device variability of the memristor's probabilistic behaviour.
ABSTRACT
Concentrations of 19 trace elements (V, Cr, Mn, Fe, Co, Cu, Zn, Se, Rb, Sr, Mo, Ag, Cd, Sb, Cs, Ba, Tl, Hg, and Pb) were determined in the liver of the striped dolphins (Stenella coeruleoalba) collected around Japan during 1977-1982 to examine the sex difference, age dependence, and interrelationships among trace elements. Tissue distribution of trace elements was also investigated in one adult and one fetus specimens. Generally, concentrations of Se, Sr, Ag, Cd, Cs, Ba, Hg, and Pb were higher in the tissues of adult than those of fetus, whereas the opposite trend was observed for Cr and Tl. There were no significant sex differences in the trace element levels in the liver. Significant positive correlations between age (0-26.5 years) and hepatic concentrations were found for Ag, Se, Hg, V, Fe, Pb, and Sr, suggesting their age-dependent accumulation in the liver. In contrast, hepatic concentrations of Mn and Zn decreased with age. Significant positive relationships were observed between Se, and Hg, Ag, V, Fe, and Sr in the liver.
Subject(s)
Liver/chemistry , Stenella/metabolism , Trace Elements/metabolism , Age Factors , Animals , Female , Fetus/chemistry , Geography , Japan , Male , Pacific Ocean , Tissue DistributionABSTRACT
Nineteen trace elements were determined in liver, muscle, kidney, gonads, and hair of 18 harp seals (Phoca groenlandica) from Pangnirtung in the Baffin Island, Canada. Concentrations of V, Mn, Fe, Cu, Mo, Ag, and Hg in the liver, Co, Cd, and Tl in the kidney, and Ba and Pb in the hair were significantly higher than those in other tissues. Significant positive correlations between Hg concentrations in the hair, and liver, kidney and testis imply usefulness of the hair sample for non-destructive monitoring of Hg in the harp seals. It is suggested that whereas Hg preferentially accumulates in the liver, the accumulation in other tissues is induced at higher hepatic Hg levels. In contrast, Se may not be accumulated in other tissues compared with the liver even at higher hepatic Hg levels because of the presence of excess Se for Hg detoxification in other tissues.