cc_lesi_oo_or.ty

"""
LESI-based orientation map with complex cell layer.

$Id: cc_lissom_oo_or.ty 8265 2008-03-26 17:09:15Z antolikjan $
"""
__version__='$Revision: 8265 $'


import numpy

from math import pi, sqrt

import param

import topo.pattern
import topo.pattern.random
import __main__

from topo.sheet.lissom import LISSOM, JointNormalizingCFSheet_Continuous
from topo.sheet.optimized import NeighborhoodMask_Opt, LISSOM_Opt
from topo.sheet import GeneratorSheet
from topo.projection import CFProjection, SharedWeightCFProjection,OneToOneProjection
from topo.responsefn.optimized import CFPRF_DotProduct_opt
from topo.base.cf import CFSheet
from topo.base.boundingregion import BoundingBox
from topo.learningfn.optimized import CFPLF_Hebbian_opt, CFPLF_Trace_opt
from topo.transferfn.optimized import CFPOF_DivisiveNormalizeL1_opt
from topo.transferfn.misc import HalfRectify
from topo.transferfn.misc import AttributeTrackingTF
from topo.transferfn import PiecewiseLinear,DivisiveNormalizeL1, HomeostaticMaxEnt,Sigmoid
from topo import numbergen
from topo.pattern import Gaussian
from topo.learningfn.projfn import CFPLF_Trace
from topo.base.functionfamily import CoordinateMapperFn
from contrib.jacommands import randomize_V1Simple_relative_LGN_strength,ActivityHysteresis, Translator,SimpleHomeoSigmoid, Jitterer, AddGC, SimpleHomeoLinear


topo.sim.name = "CCSimple"

shared_params = {"a_init":14.5, "b_init":-4, "mu":0.01, "smoothing":0.0003}
shared_debug_params = {"debug_params":["a", "b"] , "avg_params":["x"], "units":[(0,0),(11,11),(23,23)]}
core_size=locals().get('CS',0.5)

num_inputs=1

dataset=locals().get('dataset',"Natural") #set the input type by choosing the dataset parameter 


if dataset=="Natural":
        import topo.pattern.image 
        input_type=topo.pattern.image.Image
        
        image_filenames=["images/shouval/combined%02d.png"%(i+1) for i in xrange(25)]
        inputs=[input_type(filename=f,
                                size=10.0,  #size_normalization='original',(size=10.0)
                                x=0,
                                y=0,
                                orientation=0)
                        for f in image_filenames]
        input =Jitterer(generator=topo.pattern.Selector(generators=inputs),orientation=numbergen.UniformRandom(lbound=-pi,ubound=pi,seed=56),reset_period=locals().get('Period',10),jitter_magnitude=locals().get('JM',0.02))

disk_scale=locals().get('diskscale',0.35)
ring = topo.pattern.Composite(operator=numpy.add,x=numbergen.UniformRandom(lbound=-1.1,ubound=1.1,seed=12),
                                                            y=numbergen.UniformRandom(lbound=-1.1,ubound=1.1,seed=36),
                       generators=[topo.pattern.Ring(size=2, aspect_ratio=1.0, scale=0.25,thickness=locals().get('Thickness',0.01),
                                                            offset=0.0,
                                                            bounds=BoundingBox(radius=1.125), smoothing=0.03),
                                   topo.pattern.random.UniformRandom(offset=locals().get('rand_offset',0), scale=locals().get('rand_scale',0.05))])

retinal_waves=Translator(generator=ring,orientation=numbergen.UniformRandom(lbound=-pi,ubound=pi,seed=56),reset_period=locals().get('Period',10),speed=locals().get('Speed',0.02))
zeroInput = topo.pattern.Null();


jitter  =  locals().get('Jitter',0.4)
class Jitter(CoordinateMapperFn):
    """Return the jittered x,y coordinate of the given coordinate."""
    scale =  jitter
    rand = param.Parameter(default=None)
    def __call__(self,x,y):
        return x+(self.rand()-0.5)*self.scale,y+(self.rand()-0.5)*self.scale

jitterOn = Jitter(rand =numbergen.UniformRandom(seed=1023))
jitterOff = Jitter(rand =numbergen.UniformRandom(seed=1023))

# Specify weight initialization, response function, and learning function
CFProjection.weights_generator=topo.pattern.random.UniformRandom()
CFProjection.cf_shape=topo.pattern.Disk(smoothing=0.0)
CFProjection.response_fn=CFPRF_DotProduct_opt()
CFProjection.learning_fn=CFPLF_Hebbian_opt()
CFProjection.weights_output_fns=[CFPOF_DivisiveNormalizeL1_opt()]



## Homeostatic stuff
#Simple
units=[(0.25,0.25),(-0.25,0.25),
       (0.25,-0.25),(-0.25,-0.25)]

if locals().get('SOF',"Sigmoid") == "HR":
    V1Simple_OF=HalfRectify(t=locals().get('S_treshold',0.4))
elif locals().get('SOF',"Sigmoid") == "Homeo":
        V1Simple_OF = SimpleHomeoSigmoid(a_init=locals().get('a_init',22),b_init=-12,mu=locals().get('SMU',0.01),eta=locals().get('SETA',0.02))
elif locals().get('SOF',"Homeo") == "SimpleHomeo":
    V1Simple_OF = SimpleHomeoLinear(t_init=locals().get('STresh',7),alpha=locals().get('SAlpha',1),mu=locals().get('SMU',0.01),eta=locals().get('SETA',0.02))
else:
        V1Simple_OF=Sigmoid(r=22,k=-12)


if locals().get('COF',"Sigmoid") == "HR":
    V1Complex_OF=HalfRectify()
elif locals().get('COF',"Sigmoid") == "SimpleHomeo":
    V1Complex_OF = SimpleHomeoLinear(t_init=0,mu=locals().get('CMU',0.02),eta=locals().get('CETA',0.02))
else:
        V1Complex_OF=Sigmoid(r=locals().get('r_init',14.5),k=locals().get('k_init',-3))





###########################################
# build simulation

topo.sim['Retina']=GeneratorSheet(nominal_density=24.0,
                                input_generator=input,  
                                period=1.0, phase=0.05,
                                nominal_bounds=BoundingBox(radius=core_size+0.25+0.375+0.5))

topo.sim['FakeRetina1']=GeneratorSheet(nominal_density=24.0,
                                  input_generator=retinal_waves,  
                                  period=1.0, phase=0.05,
                                  nominal_bounds=BoundingBox(radius=core_size+0.25+0.5))

topo.sim['FakeRetina2']=GeneratorSheet(nominal_density=24.0,
                                  input_generator=retinal_waves,  
                                  period=1.0, phase=0.05,
                                  nominal_bounds=BoundingBox(radius=core_size+0.25+0.5))


topo.sim['LGNOn']=LISSOM(nominal_density=24.0,
                          nominal_bounds=BoundingBox(radius=core_size+0.25+0.5),
                          output_fns=[HalfRectify()],tsettle=0,
                          measure_maps=False)

topo.sim['LGNOff']=LISSOM(nominal_density=24.0,
                           nominal_bounds=BoundingBox(radius=core_size+0.25+0.5),
                           output_fns=[HalfRectify()],tsettle=0,
                           measure_maps=False)

topo.sim['V1Simple'] = JointNormalizingCFSheet_Continuous(nominal_density=locals().get('default_density',96.0),
                        nominal_bounds=BoundingBox(radius=core_size),
                        output_fns=[ActivityHysteresis(time_constant=0.35),V1Simple_OF])

                        
topo.sim['V1Complex'] = JointNormalizingCFSheet_Continuous(nominal_density=locals().get('default_density',96.0),
                        nominal_bounds=BoundingBox(radius=core_size),
                        output_fns=[ActivityHysteresis(time_constant=0.35),V1Complex_OF])

topo.sim['V1ComplexInh'] = JointNormalizingCFSheet_Continuous(nominal_density=locals().get('default_density',96.0),
                        nominal_bounds=BoundingBox(radius=core_size),
                        output_fns=[ActivityHysteresis(time_constant=0.35),V1Complex_OF])


# DoG weights for the LGN
#centerg   = Gaussian(size=0.07385,aspect_ratio=1.0,output_fns=[DivisiveNormalizeL1()])
centerg   = Gaussian(size=0.15,aspect_ratio=1.0,output_fns=[DivisiveNormalizeL1()])
surroundg = Gaussian(size=0.29540,aspect_ratio=1.0,output_fns=[DivisiveNormalizeL1()])

on_weights = topo.pattern.Composite(
    generators=[centerg,surroundg],operator=numpy.subtract)

off_weights = topo.pattern.Composite(
    generators=[surroundg,centerg],operator=numpy.subtract)


topo.sim.connect('FakeRetina1','LGNOn',delay=0.05,
                 connection_type=OneToOneProjection,strength=0.5,
                 nominal_bounds=BoundingBox(radius=0.375),name='Afferent')

topo.sim.connect('FakeRetina2','LGNOff',delay = 0.05,
                 connection_type=OneToOneProjection,strength=0.5,
                 nominal_bounds=BoundingBox(radius=0.375),name='Afferent')


g1 = Gaussian(aspect_ratio=1.0,scale=1.0,size=numbergen.UniformRandom(lbound=0.4,ubound=0.4,seed=56))
g1._Dynamic_time_fn = None
g2 = Gaussian(aspect_ratio=1.0,scale=1.0,size=numbergen.UniformRandom(lbound=0.4,ubound=0.4,seed=56))
g2._Dynamic_time_fn = None

#V1 Simple

topo.sim.connect('LGNOn','V1Simple',delay=0.05,dest_port=('Activity','JointNormalize', 'Afferent'),
                 connection_type=CFProjection,strength=locals().get('LGNStr',70),name='LGNOnAfferent',
                 weights_generator=topo.pattern.Composite(operator=numpy.multiply, 
                     generators=[g1
                 ,topo.pattern.random.UniformRandom()]),
                 nominal_bounds_template=BoundingBox(radius=0.27083),
                 coord_mapper=jitterOn,apply_output_fns_init=False,
                 learning_rate=(BoundedNumber(bounds=(0.137,None),generator=
                               ExponentialDecay(starting_value = locals().get('S_lr',5),
                                                time_constant=locals().get('S_tc',16000)))))


topo.sim.connect('LGNOff','V1Simple',delay=0.05,dest_port=('Activity','JointNormalize', 'Afferent'),
                 connection_type=CFProjection,strength=locals().get('LGNStr',70),name='LGNOffAfferent',
                 weights_generator=topo.pattern.Composite(operator=numpy.multiply, 
                     generators=[g2
                 ,topo.pattern.random.UniformRandom()]),
                 nominal_bounds_template=BoundingBox(radius=0.27083),
                 coord_mapper=jitterOff,apply_output_fns_init=False,
                 learning_rate=(BoundedNumber(bounds=(0.137,None),generator=
                               ExponentialDecay(starting_value = locals().get('S_lr',5),
                                                time_constant=locals().get('S_tc',16000)))))


simpleLateral  =  locals().get('SL',False)
if simpleLateral:
        topo.sim.connect('V1Simple','V1Simple',delay=0.05,name='LateralExcitatoryL4',
                         connection_type=CFProjection,strength=locals().get('SxLatExcStr',0.1),
                         weights_generator=topo.pattern.Gaussian(aspect_ratio=1.0, size=locals().get('SxLatExcSize',0.04)),
                         nominal_bounds_template=BoundingBox(radius=locals().get('CxLatExcBB',0.04)),
                         learning_rate=0.0)
        
        topo.sim.connect('V1Simple','V1Simple',delay=0.05,name='LateralInhibitoryL4',
                         connection_type=CFProjection,strength=-locals().get('SxLatInhStr',0.1),
                         weights_generator=topo.pattern.Composite(operator=numpy.multiply, 
                             generators=[Gaussian(aspect_ratio=1.0, size=locals().get('SxLatInhSize',0.22917)),
                                        topo.pattern.random.UniformRandom()]),
                         nominal_bounds_template=BoundingBox(radius=0.22917),
                         learning_rate=(BoundedNumber(bounds=(0.137,None),generator=
                               ExponentialDecay(starting_value = locals().get('LatInh_lr',0.3),
                                                time_constant=locals().get('LatInh_tc',16000)))))


#V1 Complex

topo.sim.connect('V1Simple','V1Complex',delay=0.05,
                 connection_type=CFProjection,strength=locals().get('StoCStr',0.4),name='V1SimpleAfferent',
                 weights_generator=Gaussian(aspect_ratio=1.0, size=locals().get('StoCSize',0.05)),
                 nominal_bounds_template=BoundingBox(radius=locals().get('StoCBB',0.15)/2),learning_rate=locals().get('FF_lr',0.0))
                
topo.sim.connect('V1Complex','V1Simple',delay=0.05,
                 connection_type=CFProjection,strength=locals().get('FStrExc',0.2),name='V1SimpleFeedbackExc1',
                 weights_generator=Gaussian(aspect_ratio=1.0, size=18),
                 nominal_bounds_template=BoundingBox(radius=locals().get('FExcSize',0.005)/2),
                 learning_rate=0)

topo.sim.connect('V1Complex','V1Simple',delay=0.05,
                 connection_type=CFProjection,strength=locals().get('FStrInh',-3.5),name='V1SimpleFeedbackInh',
                 weights_generator=Gaussian(aspect_ratio=1.0, size=2.5),
                 nominal_bounds_template=BoundingBox(radius=locals().get('FInhSize',0.3)/2),learning_rate=0)


topo.sim.connect('V1Complex','V1Complex',delay=0.05,name='LateralExcitatoryToExcitatory',
                 connection_type=CFProjection,strength=locals().get('CxLatExcExcStr',2.0),
                 weights_generator=topo.pattern.Gaussian(aspect_ratio=1.0, size=locals().get('CxLatExcSize',0.06)),
                 nominal_bounds_template=BoundingBox(radius=locals().get('CxLatExcBB',0.25)),
                 learning_rate=locals().get('lat_exc_lr',0.0))

topo.sim.connect('V1Complex','V1ComplexInh',delay=0.05,name='LateralExcitatoryToInhibitory',
                 connection_type=CFProjection,strength=locals().get('CxLatExcInhStr',1.5),
                 weights_generator=topo.pattern.Gaussian(aspect_ratio=1.0, size=locals().get('CxLatExcSize',0.06)),
                 nominal_bounds_template=BoundingBox(radius=locals().get('CxLatExcBB',0.25)),
                 learning_rate=locals().get('lat_exc_lr',0.0))

topo.sim.connect('V1ComplexInh','V1Complex',delay=0.05,name='LateralInhibitoryToExcitatory',
                 connection_type=CFProjection,strength=-locals().get('CxLatInhExcStr',2.0),
                 weights_generator=topo.pattern.Gaussian(aspect_ratio=1.0, size=locals().get('CxLatInhSize',0.12)),
                 nominal_bounds_template=BoundingBox(radius=locals().get('CxLatInhBB',0.12)),
                 learning_rate=locals().get('lat_inh_lr',0.0))

topo.sim.connect('V1ComplexInh','V1ComplexInh',delay=0.05,name='LateralInhibitoryToInhibitory',
                 connection_type=CFProjection,strength=-locals().get('CxLatInhInhStr',1.3),
                 weights_generator=topo.pattern.Gaussian(aspect_ratio=1.0, size=locals().get('CxLatInhSize',0.12)),
                 nominal_bounds_template=BoundingBox(radius=locals().get('CxLatInhBB',0.12)),
                 learning_rate=locals().get('lat_inh_lr',0.0))


topo.sim.schedule_command(__main__.__dict__.get('SST',10000),"secondStage()")

AddGC()


def secondStage():
    topo.sim.connect('Retina','LGNOn',delay=0.05,
                    connection_type=SharedWeightCFProjection,strength=2.33,
                    nominal_bounds_template=BoundingBox(radius=0.375),name='LGNOnAfferent1',
                    weights_generator=on_weights)
    
    topo.sim.connect('Retina','LGNOff',delay = 0.05,
                    connection_type=SharedWeightCFProjection,strength=2.33,
                    nominal_bounds_template=BoundingBox(radius=0.375),name='LGNOffAfferent1',
                    weights_generator=off_weights)
    
    topo.sim['FakeRetina1'].set_input_generator(zeroInput)
    topo.sim['FakeRetina2'].set_input_generator(zeroInput)
    topo.sim['LGNOn'].in_connections[0].strength=0
    topo.sim['LGNOff'].in_connections[0].strength=0
    randomize_V1Simple_relative_LGN_strength(prob=__main__.__dict__.get('CRS',0.5))
    
from topo.analysis.featureresponses import MeasureResponseCommand
MeasureResponseCommand.scale=3    

def check_activity(a,b,c,d,e,f,g,h):
    print a,b,c,d,e,f,g

    import pylab
    prefix="/disk/scratch/ActivityExploration1/"
    
    pylab.figure(1)

    V1Splastic =     topo.sim["V1Simple"].plastic
    V1Cplastic =     topo.sim["V1Complex"].plastic
    topo.sim["V1Simple"].plastic = False
    topo.sim["V1Complex"].plastic = False
    
    topo.sim.state_push()
   
    topo.sim["V1Simple"].in_connections[0].strength=a
    topo.sim["V1Simple"].in_connections[0].strength=a
    
    topo.sim["V1Simple"].projections()["V1SimpleFeedbackExc1"].strength=b
    topo.sim["V1Simple"].projections()["V1SimpleFeedbackInh"].strength=c
    topo.sim["V1Complex"].projections()["LateralExcitatoryToExcitatory"].strength=d
    topo.sim["V1ComplexInh"].projections()["LateralExcitatoryToInhibitory"].strength=e
    topo.sim["V1Complex"].projections()["LateralInhibitoryToExcitatory"].strength=f
    topo.sim["V1ComplexInh"].projections()["LateralInhibitoryToInhibitory"].strength=g
    topo.sim["V1Complex"].projections()["V1SimpleAfferent"].strength=h
    try:
        topo.sim.run(0.8)
        activity = topo.sim["V1Complex"].activity
        topo.sim.run(0.05)
        activity += topo.sim["V1Complex"].activity
        topo.sim.run(0.05)
        activity += topo.sim["V1Complex"].activity
        topo.sim.run(0.05)
        activity += topo.sim["V1Complex"].activity
        topo.sim.run(0.05)
        activity += topo.sim["V1Complex"].activity
        
        pylab.clf()
        pylab.imshow(activity)
        pylab.colorbar()
        pylab.savefig(prefix+ "Activity:"  + "_" + str(a)+ "_" + str(b) + "_" + str(c) + "_" + str(d)+ "_" + str(e)  + "_" + str(f) + "_" + str(g) +".png");
    except FloatingPointError:
        print "Error"
        pass

    topo.sim.state_pop()
    
    topo.sim["V1Simple"].plastic = V1Splastic 
    topo.sim["V1Complex"].plastic = V1Cplastic
    
    
#contrib.jacommands.run_combinations(check_activity,[[70,60,80],[0.2,0.15,0.25],[-3.5,-3.0,-4.0],[2.0,1.5,2.5],[1.5,1.0,2.0],[-2.0,-1.5,-2.5],[-1.0,-0.8,-1.2],[0.4,0.3,0.5]])