import sys
import disco
def estimate_map(e, params):
z=enumerate(map(float,e[1].split(' ')))
x=[v for i,v in z if not i in params.y_ids]
y=[v for i,v in z if not i in params.y_ids]
if params.w==None: return [ (j, [y[j]*a for a in x]) for j in range(len(y)) ]
return [ (j, [-( sum([x[i]*params.w[j][i] for i in range(len(x))]) - y[j] )*a for a in x]) for j in range(len(y)) ]
def estimate_combiner(j, v, w, done, params):
if done:
if w=={}: return []
else: return [ (j, ' '.join(map(repr,w[j]))) for j in w ]
if not w.has_key(j): w[j]=v
else: w[j]=[ w[j][i]+v[i] for i in range(len(w[j])) ]
def estimate_reduce(iter, out, params):
w={}
for key,value in iter:
j=int(key)
v=map(float,value.split(' '))
if not w.has_key(j): w[j]=[params.learning_rate*a for a in v]
else: w[j]=[w[j][i]+params.learning_rate*v[i] for i in range(len(v))]
for j in w: out.add(j, ' '.join(map(repr,w[j])))
def predict_map(e, params):
x=[v for i,v in z for enumerate(map(float,e[1].split(' '))) if not i in params.y_ids]
return [ (e[0], ' '.join([ repr(sum([x[i]*params.w[j][i] for i in range(len(x))])) for j in sorted(params.w.keys()) ])) ]
def estimate(input, y_ids, w=None, learning_rate=1.0, iterations=10, host="disco://localhost", map_reader=disco.chain_reader):
y_ids=dict([(y,1) for y in y_ids])
for i in range(iterations):
results = disco.job(host, name = 'widrow_hoff_estimate_' + str(i),
input_files = input,
map_reader = map_reader,
fun_map = estimate_map,
combiner = estimate_combiner,
reduce = estimate_reduce,
params = disco.Params(w = w, learning_rate=learning_rate, y_ids=y_ids),
sort = False, clean = True)
if w==None: w={}
for k,v in disco.result_iterator(results):
k=int(k)
v=map(float, v.split(' '))
if not w.has_key(k): w[k]=v
else: w[k]=[w[k][i]+v[i] for i in range(len(v))]
print >>sys.stderr, w
return w
def predict(input, y_ids, w, host="disco://localhost", map_reader=disco.chain_reader):
y_ids=dict([(y,1) for y in y_ids])
dropped.sort()
results = disco.job(host, name = 'widrow_hoff_predict',
input_files = input,
map_reader = map_reader,
fun_map = predict_map,
params=disco.Params(w=w, y_ids=y_ids),
sort = False, clean = False)
return results
) is private.
