Journal: Journal of Neurophysiology

Article Title: A latent low-dimensional common input drives a pool of motor neurons: a probabilistic latent state-space model

doi: 10.1152/jn.00274.2017

Figure Lengend Snippet: The state-space model represents the processes that are engaged to enable a population of motor units to produce a force that matches a desired target force. The trajectory of the latent state, X(t), must evolve to provide sufficient activation to the pool of motor neurons for them to discharge action potentials. The changes in the latent state over time are captured by the matrix, A, which represents a linear approximation of the latent-state dynamics. All neurons receive two inputs, a common input (the latent state) that excites the motor neurons and zero-mean Gaussian synaptic noise. The physiological properties of each neuron are modeled with the C matrix, which encodes the effects of the input signal on discharge rate. The output is instantaneous discharge rates of each motor neuron, which evolve as the latent state changes. The forces generated by the activated motor units are summed to create the net muscle force.

Article Snippet: In contrast, the state-space model estimated a trajectory of the latent state with a higher frequency content at 4 Hz compared with 3 Hz for all three current amplitudes ( , C , F , and I ), which indicates greater sensitivity for the state-space model to the details of the input currents. fig ft0 fig mode=article f1 fig/graphic|fig/alternatives/graphic mode="anchored" m1 Open in a separate window Fig. 7. caption a7 Ten integrate-and-fire neurons were simulated when receiving input currents at 2 frequencies and 3 amplitudes for 10 s. A , D , and G : the common input current at 3 (gray) and 4 Hz (black) with superimposed Gaussian noise (awgn with a signal-to-noise ratio of 25 in MATLAB) used to activate the 10 neurons.

Techniques: Activation Assay, Generated

Journal: Journal of Neurophysiology

Article Title: A latent low-dimensional common input drives a pool of motor neurons: a probabilistic latent state-space model

doi: 10.1152/jn.00274.2017

Figure Lengend Snippet: Estimation of the state-space model from a pair of integrate-and-fire neurons (see Fig. 2) simulated for 1 (A and B) and 10 s (C and D). Both neurons discharged action potentials in response to a 0.1-nA sinusoidal current with an input frequency of 3 Hz as well as receiving independent and common Gaussian noise. Additionally, neuron B received an inhibitory current from neuron C (a Renshaw cell) in response to input from neuron A, which reduced the discharge rate of neuron B. The trajectory of the single-dimension approximation of the latent state for the pair of neurons as solved with the EM algorithm is shown as a black solid line (Buesing et al. 2012; Macke et al. 2011), the injected current with common noise superimposed is drawn as a dashed line, and the low-pass-filtered (10 Hz) cumulative spike train (CST) is indicated as a gray line. There was a strong temporal relation between peaks in the injected current and changes in the latent state for both simulations (1 and 10 s), whereas the association between injected current and modulation of CST was observed only for the longer simulation (C). There was moderate coherence (0.58) between the injected current and latent state at ∼2.5 Hz (data not shown). B and D: coherence between the low-pass-filtered CST and the input current. The coherence was moderate for all low-frequency (<10 Hz) values due to the inhibitory input from the Renshaw cell (C) onto neuron B as well as the brief length of the trial. However, coherence between the 2 signals was much stronger for the 10-s simulation (D).

Article Snippet: In contrast, the state-space model estimated a trajectory of the latent state with a higher frequency content at 4 Hz compared with 3 Hz for all three current amplitudes ( , C , F , and I ), which indicates greater sensitivity for the state-space model to the details of the input currents. fig ft0 fig mode=article f1 fig/graphic|fig/alternatives/graphic mode="anchored" m1 Open in a separate window Fig. 7. caption a7 Ten integrate-and-fire neurons were simulated when receiving input currents at 2 frequencies and 3 amplitudes for 10 s. A , D , and G : the common input current at 3 (gray) and 4 Hz (black) with superimposed Gaussian noise (awgn with a signal-to-noise ratio of 25 in MATLAB) used to activate the 10 neurons.

Techniques: Injection

Journal: Journal of Neurophysiology

Article Title: A latent low-dimensional common input drives a pool of motor neurons: a probabilistic latent state-space model

doi: 10.1152/jn.00274.2017

Figure Lengend Snippet: Ten integrate-and-fire neurons were simulated when receiving input currents at 2 frequencies and 3 amplitudes for 10 s. A, D, and G: the common input current at 3 (gray) and 4 Hz (black) with superimposed Gaussian noise (awgn with a signal-to-noise ratio of 25 in MATLAB) used to activate the 10 neurons. B, E, and H: spectral content of the low-pass-filtered CST derived with the Fast Fourier transform (FFT) function in MATLAB. C, F, and I: estimation of the latent-state trajectory from the discharge activity of the 10 integrate-and-fire neurons at the 2 frequencies and 3 input-current amplitudes.

Article Snippet: In contrast, the state-space model estimated a trajectory of the latent state with a higher frequency content at 4 Hz compared with 3 Hz for all three current amplitudes ( , C , F , and I ), which indicates greater sensitivity for the state-space model to the details of the input currents. fig ft0 fig mode=article f1 fig/graphic|fig/alternatives/graphic mode="anchored" m1 Open in a separate window Fig. 7. caption a7 Ten integrate-and-fire neurons were simulated when receiving input currents at 2 frequencies and 3 amplitudes for 10 s. A , D , and G : the common input current at 3 (gray) and 4 Hz (black) with superimposed Gaussian noise (awgn with a signal-to-noise ratio of 25 in MATLAB) used to activate the 10 neurons.

Techniques: Derivative Assay, Activity Assay

Journal: Entropy

Article Title: Enhanced Partial Discharge Signal Denoising Using Dispersion Entropy Optimized Variational Mode Decomposition

doi: 10.3390/e23121567

Figure Lengend Snippet: The image on the ( left ) shows synthetic ’Bumps’ signal of 1024 samples and the image on ( right ) shows noisy ‘Bumps’ signal with added white Gaussian noise of 10 dB.

Article Snippet: The MATLAB ® built-in function ‘ awgn ’ is used to add white Gaussian noise with a Signal-to-Noise Ratio (SNR) of −5 dB, 0 dB, 5 dB, 10 dB and 20 dB.

Techniques: