"""
Specific support for the LISSOM algorithm.
"""
import param
import topo
from topo.sheet import SettlingCFSheet
from topo.transferfn import PiecewiseLinear
# Legacy declaration for backwards compatibility
class LISSOM(SettlingCFSheet):
output_fns = param.HookList(default=[PiecewiseLinear(lower_bound=0.1,upper_bound=0.65)])
[docs]def schedule_events(sheet_str="topo.sim['V1']",st=0.5,aff_name="Afferent",
ids=1.0,ars=1.0,increase_inhibition=False):
"""
Convenience function for scheduling a default set of events
typically used with a LISSOM model sheet. The parameters used
are the defaults from Miikkulainen, Bednar, Choe, and Sirosh
(2005), Computational Maps in the Visual Cortex, Springer.
Note that Miikulainen 2005 specifies only one output_fn for the
LISSOM model sheet; where these scheduled actions operate on an
output_fn, they do so only on the first output_fn in the sheet's
list of output_fns.
Installs afferent learning rate changes for any projection whose
name contains the keyword specified by aff_name (typically
"Afferent").
The st argument determines the timescale relative to a
20000-iteration simulation, and results in the default
10000-iteration simulation for the default st=0.5.
The ids argument specifies the input density scale, i.e. how much
input there is at each iteration, on average, relative to the
default. The ars argument specifies how much to scale the
afferent learning rate, if necessary.
If increase_inhibition is true, gradually increases the strength
of the inhibitory connection, typically used for natural image
simulations.
"""
# Allow sheet.BoundingBox calls (below) after reloading a snapshot
topo.sim.startup_commands.append("from topo import sheet")
# Lateral excitatory bounds changes
# Convenience variable: excitatory projection
LE=sheet_str+".projections()['LateralExcitatory']"
topo.sim.schedule_command( 200*st,LE+'.change_bounds(sheet.BoundingBox(radius=0.06250))')
topo.sim.schedule_command( 500*st,LE+'.change_bounds(sheet.BoundingBox(radius=0.04375))')
topo.sim.schedule_command( 1000*st,LE+'.change_bounds(sheet.BoundingBox(radius=0.03500))')
topo.sim.schedule_command( 2000*st,LE+'.change_bounds(sheet.BoundingBox(radius=0.02800))')
topo.sim.schedule_command( 3000*st,LE+'.change_bounds(sheet.BoundingBox(radius=0.02240))')
topo.sim.schedule_command( 4000*st,LE+'.change_bounds(sheet.BoundingBox(radius=0.01344))')
topo.sim.schedule_command( 5000*st,LE+'.change_bounds(sheet.BoundingBox(radius=0.00806))')
topo.sim.schedule_command( 6500*st,LE+'.change_bounds(sheet.BoundingBox(radius=0.00484))')
topo.sim.schedule_command( 8000*st,LE+'.change_bounds(sheet.BoundingBox(radius=0.00290))')
topo.sim.schedule_command(20000*st,LE+'.change_bounds(sheet.BoundingBox(radius=0.00174))')
# Lateral excitatory learning rate changes
idss=("" if ids==1 else "/%3.1f"%ids)
estr='%s.learning_rate=%%s%s*%s.n_units'%(LE,idss,LE)
topo.sim.schedule_command( 200*st,estr%'0.12168')
topo.sim.schedule_command( 500*st,estr%'0.06084')
topo.sim.schedule_command( 1000*st,estr%'0.06084')
topo.sim.schedule_command( 2000*st,estr%'0.06084')
topo.sim.schedule_command( 3000*st,estr%'0.06084')
topo.sim.schedule_command( 4000*st,estr%'0.06084')
topo.sim.schedule_command( 5000*st,estr%'0.06084')
topo.sim.schedule_command( 6500*st,estr%'0.06084')
topo.sim.schedule_command( 8000*st,estr%'0.06084')
topo.sim.schedule_command(20000*st,estr%'0.06084')
### Lateral inhibitory learning rate and strength changes
if increase_inhibition:
LI=sheet_str+".projections()['LateralInhibitory']"
istr='%s.learning_rate=%%s%s'%(LI,idss)
topo.sim.schedule_command( 1000*st,istr%'1.80873/5.0*2.0')
topo.sim.schedule_command( 2000*st,istr%'1.80873/5.0*3.0')
topo.sim.schedule_command( 5000*st,istr%'1.80873/5.0*5.0')
topo.sim.schedule_command( 1000*st,LI+'.strength=-2.2')
topo.sim.schedule_command( 2000*st,LI+'.strength=-2.6')
# Afferent learning rate changes (for every Projection named Afferent)
sheet_=eval(sheet_str)
projs = [pn for pn in sheet_.projections().keys() if pn.count(aff_name)]
num_aff=len(projs)
arss="" if ars==1.0 else "*%3.1f"%ars
for pn in projs:
ps="%s.projections()['%s'].learning_rate=%%s%s%s" % \
(sheet_str,pn,idss if num_aff==1 else "%s/%d"%(idss,num_aff),arss)
topo.sim.schedule_command( 500*st,ps%('0.6850'))
topo.sim.schedule_command( 2000*st,ps%('0.5480'))
topo.sim.schedule_command( 4000*st,ps%('0.4110'))
topo.sim.schedule_command(20000*st,ps%('0.2055'))
# Activation function threshold changes
bstr = sheet_str+'.output_fns[0].lower_bound=%5.3f;'+\
sheet_str+'.output_fns[0].upper_bound=%5.3f'
lbi=sheet_.output_fns[0].lower_bound
ubi=sheet_.output_fns[0].upper_bound
topo.sim.schedule_command( 200*st,bstr%(lbi+0.01,ubi+0.01))
topo.sim.schedule_command( 500*st,bstr%(lbi+0.02,ubi+0.02))
topo.sim.schedule_command( 1000*st,bstr%(lbi+0.05,ubi+0.03))
topo.sim.schedule_command( 2000*st,bstr%(lbi+0.08,ubi+0.05))
topo.sim.schedule_command( 3000*st,bstr%(lbi+0.10,ubi+0.08))
topo.sim.schedule_command( 4000*st,bstr%(lbi+0.10,ubi+0.11))
topo.sim.schedule_command( 5000*st,bstr%(lbi+0.11,ubi+0.14))
topo.sim.schedule_command( 6500*st,bstr%(lbi+0.12,ubi+0.17))
topo.sim.schedule_command( 8000*st,bstr%(lbi+0.13,ubi+0.20))
topo.sim.schedule_command(20000*st,bstr%(lbi+0.14,ubi+0.23))
# Just to get more progress reports
topo.sim.schedule_command(12000*st,'pass')
topo.sim.schedule_command(16000*st,'pass')
# Settling steps changes
topo.sim.schedule_command( 2000*st,sheet_str+'.tsettle=10')
topo.sim.schedule_command( 5000*st,sheet_str+'.tsettle=11')
topo.sim.schedule_command( 6500*st,sheet_str+'.tsettle=12')
topo.sim.schedule_command( 8000*st,sheet_str+'.tsettle=13')
__all__ = [
"schedule_events",
]
