kuhess/pso-ann

Particle Swarm Optimization (PSO) to optimize Artificial Neural Network (ANN)

pso-ann's Issues

BUG pso.py

I think the code
‘self.V = self.chi * （self.V
+ self.phi_p * R_p * (self.P - self.X)
+ self.phi_g * R_g * (self.g - self.X)）’
should be
‘self.V = self.chi * self.V
+ self.phi_p * R_p * (self.P - self.X)
+ self.phi_g * R_g * (self.g - self.X)’.

How to interpret the weigths

Once you have optimized the weights for ann, how can I interpret the acquired weigths so I can get the connections between nodes?

BUG example_mnist.py

Hi I believe there is a bug in 'example_mnist.py' in the function
vector_to_weights.
The indexing variable idx should be updated after the first iteration s.t.
idx = idx_max

Otherwise not the whole vector of weights is read.

bug of pso.py

In pso.py, the updated velocity is written in Lines 65-69 as follows:
self.V = self.chi * (
self.V
+ self.phi_p * R_p * (self.P - self.X)
+ self.phi_g * R_g * (self.g - self.X)
)
however, it is written on Wikipedia
vi,d ← wvi,d + φprp*(pi,d-xi,d) + φgrg(gd-xi,d)
Does it delete the bracket and change to this code?
self.V = self.chi * self.V + self.phi_p * R_p * (self.P - self.X) + self.phi_g * R_g * (self.g - self.X)

some questions in ann

hi,I am Senior in a university of china and I am stduying about pso-ann model for wifi-location indoor.
I hava read your code and the output of the digital-data is a single number, the output-layer just one node. however in my research,the ann out put should be a coordinate（x,y）,means two output，so could u please show me how to modify the model to meet the output. thanks a lot.
my email : [email protected]