Clarify in docs that Poisson noise is generated using spike_generator not poisson_generator

Author: akorgor

pynest/examples/eprop_plasticity/eprop_supervised_classification_evidence-accumulation_bsshslm_2020.py

@@ -47,7 +47,7 @@
 
 Learning in the neural network model is achieved by optimizing the connection weights with e-prop plasticity.
 This plasticity rule requires a specific network architecture depicted in Figure 1. The neural network model
-consists of a recurrent network that receives input from Poisson generators and projects onto two readout
+consists of a recurrent network that receives input from spike generators and projects onto two readout
 neurons - one for the left and one for the right turn at the end. The input neuron population consists of four
 groups: one group providing background noise of a specific rate for some base activity throughout the
 experiment, one group providing the input spikes of the left cues and one group providing them for the right

pynest/examples/eprop_plasticity/eprop_supervised_classification_neuromorphic_mnist.py

@@ -44,7 +44,7 @@
 
 Learning in the neural network model is achieved by optimizing the connection weights with e-prop plasticity.
 This plasticity rule requires a specific network architecture depicted in Figure 1. The neural network model
-consists of a recurrent network that receives input from Poisson generators and projects onto multiple readout
+consists of a recurrent network that receives input from spike generators and projects onto multiple readout
 neurons - one for each class. Each input generator is assigned to a pixel of the input image; when an event is
 detected in a pixel at time :math:`t`, the corresponding input generator (connected to an input neuron) emits a spike
 at that time. Each readout neuron compares the network signal :math:`y_k` with the teacher signal :math:`y_k^*`,
@@ -177,10 +177,10 @@
 # We proceed by creating a certain number of input, recurrent, and readout neurons and setting their parameters.
 # Additionally, we already create an input spike generator and an output target rate generator, which we will
 # configure later. Each input sample is mapped out to a 34x34 pixel grid and a polarity dimension. We allocate
-# Poisson generators to each input image pixel to simulate spike events. However, due to the observation
+# spike generators to each input image pixel to simulate spike events. However, due to the observation
 # that some pixels either never record events or do so infrequently, we maintain a blocklist of these inactive
-# pixels. By omitting Poisson generators for pixels on this blocklist, we effectively reduce the total number of
-# input neurons and Poisson generators required, optimizing the network's resource usage.
+# pixels. By omitting spike generators for pixels on this blocklist, we effectively reduce the total number of
+# input neurons and spike generators required, optimizing the network's resource usage.
 
 pixels_blocklist = np.loadtxt("./NMNIST_pixels_blocklist.txt")
 

pynest/examples/eprop_plasticity/eprop_supervised_regression_handwriting_bsshslm_2020.py

@@ -45,7 +45,7 @@
 
 Learning in the neural network model is achieved by optimizing the connection weights with e-prop plasticity.
 This plasticity rule requires a specific network architecture depicted in Figure 1. The neural network model
-consists of a recurrent network that receives frozen noise input from Poisson generators and projects onto two
+consists of a recurrent network that receives frozen noise input from spike generators and projects onto two
 readout neurons. Each individual readout signal denoted as :math:`y_k` is compared with a corresponding target
 signal represented as :math:`y_k^*`. The network's training error is assessed by employing a mean-squared error
 loss.

pynest/examples/eprop_plasticity/eprop_supervised_regression_lemniscate_bsshslm_2020.py

@@ -45,7 +45,7 @@
 
 Learning in the neural network model is achieved by optimizing the connection weights with e-prop plasticity.
 This plasticity rule requires a specific network architecture depicted in Figure 1. The neural network model
-consists of a recurrent network that receives frozen noise input from Poisson generators and projects onto two
+consists of a recurrent network that receives frozen noise input from spike generators and projects onto two
 readout neurons. Each individual readout signal denoted as :math:`y_k` is compared with a corresponding target
 signal represented as :math:`y_k^*`. The network's training error is assessed by employing a mean-squared error
 loss.

pynest/examples/eprop_plasticity/eprop_supervised_regression_sine-waves.py

@@ -46,7 +46,7 @@
 
 Learning in the neural network model is achieved by optimizing the connection weights with e-prop plasticity.
 This plasticity rule requires a specific network architecture depicted in Figure 1. The neural network model
-consists of a recurrent network that receives frozen noise input from Poisson generators and projects onto one
+consists of a recurrent network that receives frozen noise input from spike generators and projects onto one
 readout neuron. The readout neuron compares the network signal :math:`y` with the teacher target signal
 :math:`y*`, which it receives from a rate generator. In scenarios with multiple readout neurons, each individual
 readout signal denoted as :math:`y_k` is compared with a corresponding target signal represented as

pynest/examples/eprop_plasticity/eprop_supervised_regression_sine-waves_bsshslm_2020.py

@@ -45,7 +45,7 @@
 
 Learning in the neural network model is achieved by optimizing the connection weights with e-prop plasticity.
 This plasticity rule requires a specific network architecture depicted in Figure 1. The neural network model
-consists of a recurrent network that receives frozen noise input from Poisson generators and projects onto one
+consists of a recurrent network that receives frozen noise input from spike generators and projects onto one
 readout neuron. The readout neuron compares the network signal :math:`y` with the teacher target signal
 :math:`y*`, which it receives from a rate generator. In scenarios with multiple readout neurons, each individual
 readout signal denoted as :math:`y_k` is compared with a corresponding target signal represented as