Recently, with the broad applications of deep learning technology in image denoising, many deep neural networks based on the residual block (ResBlock) have been proposed to improve the performance. This paper presents DC-ResBlock, a ResBlock with an extra dilated convolution, to replace standard ResBlock for image denoising. Our study shows that the deep neural network with DC-ResBlocks, named DRNet, can achieve a very competitive result. We experiment with other residual blocks by adding Rectified Linear Unit, batch normalization, and Dilated Convolution to the standard ResBlock in several different ways. Ablation study shows that DC-Resblock has the best denoising performance. To evaluate the robustness of proposed DRNet, we make a statistical measure using many different random seeds instead of just a single seed, as used by many previous studies. DRNet performs well in denoising gray and color images with additive white Gaussian noise. Furthermore, because of its good performance on the Smartphone Image Denoising Dataset, DRNet is hopeful to be applied in practical tasks. The code of DRNet is accessible at https://github.com/JiaHongZ/DRNet.
The formation and consolidation of memory play a vital role for survival in an ever-changing environment. In the brain, the change and stabilization of potentiated and depressed synapses are the neural basis of memory formation and maintenance. These changes can be induced by rather short stimuli (only a few seconds or even less) but should then be stable for months or years. Recently, the neural mechanism of conversion from rapid change during the early phase of synaptic plasticity into a stable memory trace in the late phase of synaptic plasticity is more and more clear at the protein and molecular levels, among which synaptic tagging and capture (STC) theory is one of the most popular theories. According to the STC theory, the change and stabilization of synaptic efficiency mainly depend on three processes related to calcium concentration, including synaptic tagging, synthesis of plasticity-related product (PRP), and the capture of PRP by tagged synapse. Based on the STC theory, several computational models are proposed. However, these models hardly take simplicity and biological interpretability into account simultaneously. Here, we propose a simplified STC (SM-STC) model to address this issue. In the SM-STC model, the concentration of calcium ion in each neuronal compartment and synapse is first calculated, and then the tag state of synapse and PRP are updated, and the coupling effect of tagged synapse and PRP is further considered to determine the plasticity state of the synapse, either potentiation or depression. We simulated the Schaffer collaterals pathway of the hippocampus targeting a multicompartment CA1 neuron for several hours of biological time. The results show that the SM-STC model can produce a broad range of experimental phenomena known in the physiological experiments, including long-term potentiation induced by high-frequency stimuli, long-term depression induced by low-frequency stimuli, and cross-capture with two stimuli separated by a delay. Thus, the SM-STC model proposed in this study provides an effective learning rule for brain-like computation on the premise of ensuring biological plausibility and computational efficiency.
7 February 2022 in Frontiers in Computational Neuroscience
Hodgkin-Huxley (HH)-type model is the most famous computational model for simulating neural activity. It shows the highest accuracy in capturing neuronal spikes, and its model parameters have definite physiological meanings. However, HH-type models are computationally expensive. To address this problem, a previous study proposed a spike prediction module (SPM) to predict whether a spike will take place 1 ms later based on three voltage values with intervals of 1 ms. Although SPM does well, it fails to evaluate the informative features of the spike. In this study, the feature prediction module (FPM) based on simple artificial neural network (ANN) was proposed to predict spike features including maximum voltage, minimum voltage, and dropping interval. Nine different HH-type models were adopted whose firing patterns cover most of the firing behaviors observed in the brain. Voltage and spike feature samples under constant external input current were collected for training and testing. Experiment results illustrated that the combination of SPM and FPM can accurately predict the spiking part of different HH-type models and can generalize to unseen types of input current. The combination of SPM and FPM may offer a possible way to simulate the action potentials of biological neurons with high accuracy and efficiency.