FSD for RRG

Basic information

Time Series Plot represents average membrane potential versus iteration number.

(1)

\begin{align} \mbox{avg membr. pot.}(t) = \frac{ \sum_{i=1}^{n} V_i(t) } { n } \end{align}

where $V_i(t)$ is membrane potential of $ith$ neuron in time moment $t$ and $n$ is number of neurons in network.

Network consists of $n=50$ follower cells.

Time series from RRG with one burst

---

Time series information

simulation time interval = 0.01
saving to file every: 1 iteration
sample size: 49090

---

Time series plot

---

Trajectories for 4 clusters

Gauss dispersions: $\frac{1}{2} \left\| \mu_i - \mu_j \right\|$

---

Animation of trajectory (2 best clusters) and time series

2 best clusters means clusters that k-means algorithm finds for $k = 2$ parameter. In every start of k-means algorithm these 2 clusters were found.

Gauss dispersions: $\frac{1}{2} \left\| \mu_1 - \mu_2 \right\|$

---

Static version of previous trajectory with Gauss centers comparison

---

Neuron vs neuron activities plot

Neuronal activities plot of 2 neurons that have the most (left) and the least (right) different vectors of neuronal activities.

Time series from RRG with 5 bursts

---

Time series information

simulation time interval = 0.05
save to file every: 1 iterations
sample size: 49455

---

Time series plot

---

Trajectories for 4 clusters

Gauss dispersions: $\frac{1}{2} \left\| \mu_i - \mu_j \right\|$

Time series from RRG with 6 bursts - pathological case

---

Time series information

simulation time interval = 0.05
save to file every: 1 iterations
sample size: 50000

---

Time series plot

---

Trajectories for 4 clusters

Gauss dispersions: $\frac{1}{2} \left\| \mu_i - \mu_j \right\|$

Comparison of time series for one burst, 5 bursts and pathological case

For each case 2 clusters with k-means algorithm were found.
Trajectory lines were bolded such that only the global structure is visible.

Gauss dispersions: $\frac{1}{2} \left\| \mu_i - \mu_j \right\|$

From top: one burst, 5 bursts and pathological case.
Click image to enlarge.

Comparison of time series with many bursts

Time series information

simulation time interval = 0.05
saving to file every: 10 iteration
sample size: 19600 (for each)

---

Time series and trajectory plots

For each case 2 clusters with k-means algorithm were found.
Trajectory lines were bolded such that only the global structure is visible.

Gauss dispersions: $\frac{1}{2} \left\| \mu_i - \mu_j \right\|$

On the top normal case and on the bottom pathological one.
Click image to enlarge.