def exotic_data_gen(ppc=300):
#generate concentric rings
#central: gaussian at 0,0
#ring: sample from [-pi,pi] uniformly
#sample from [-r,r] normally
xvals = np.random.uniform(-5, 5, size=(ppc, ))
deviations = np.random.normal(0.0, 0.1, size=(ppc, )).reshape((-1, 1))
yvals = ((25 - xvals**2)**0.5)
yvals2 = -yvals
locs = np.array(np.ones((ppc, 2)))
locs2 = np.ones((ppc, 2))
locs[:, 0] = xvals
locs[:, 1] = yvals
locs2[:, 0] = xvals
locs2[:, 1] = yvals2
#move the random deviations
locs = locs + deviations * locs
locs2 = locs2 + deviations * locs2
eye = ndist([0.0, 0.0], [[0.1, 0], [0, 0.1]], (ppc))
X = np.concatenate([locs, locs2, eye])
y = np.concatenate(
[np.zeros((ppc, )),
np.ones((ppc, )),
np.ones((ppc, )) * 2])
return X, y
def exotic_data_gen(ppc=300): #generate concentric rings #central: gaussian at 0,0 #ring: sample from [-pi,pi] uniformly #sample from [-r,r] normally xvals = np.random.uniform(-5,5,size=(ppc,)) deviations = np.random.normal(0.0,0.1,size=(ppc,)).reshape((-1,1)) yvals = ((25-xvals**2)**0.5) yvals2 = -yvals locs = np.array(np.ones((ppc,2))) locs2 = np.ones((ppc,2)) locs[:,0] = xvals locs[:,1] = yvals locs2[:,0] = xvals locs2[:,1] = yvals2 #move the random deviations locs = locs + deviations*locs locs2 = locs2 + deviations*locs2 eye = ndist([0.0,0.0],[[0.1,0],[0,0.1]],(ppc)) X = np.concatenate([locs,locs2,eye]) y = np.concatenate([np.zeros((ppc,)),np.ones((ppc,)),np.ones((ppc,))*2]) return X,y
def gaussian_data_gen(points_per_class=500):
#generate some data for model evaluation/building
ppc = points_per_class
sig = np.array([[.3, 0], [0, .3]])
#cents = [[-2,0],[2,0],[0,2],[0,-2]]
cents = [[-2, 0], [0, 2], [2, 0]]
points, targets = [], []
for i, c in enumerate(cents):
targets.append(np.ones((ppc, )) * i)
points.append(ndist(c, sig, (ppc)))
X = np.concatenate(points, axis=0)
y = np.concatenate(targets, axis=0)
inds = range(X.shape[0])
np.random.shuffle(inds)
Xshuf = []
yshuf = []
for i in inds:
Xshuf.append(X[i])
yshuf.append(y[i])
Xshuf = np.array(Xshuf)
yshuf = np.array(yshuf)
#TODO: handle the bimodal case.
return Xshuf, yshuf
def gaussian_data_gen(points_per_class=500): #generate some data for model evaluation/building ppc = points_per_class sig = np.array([[.3,0],[0,.3]]) #cents = [[-2,0],[2,0],[0,2],[0,-2]] cents= [[-2,0],[0,2],[2,0]] points,targets = [],[] for i,c in enumerate(cents): targets.append(np.ones((ppc,))*i) points.append(ndist(c,sig,(ppc))) X = np.concatenate(points,axis=0) y = np.concatenate(targets,axis=0) inds = range(X.shape[0]) np.random.shuffle(inds) Xshuf = [] yshuf = [] for i in inds: Xshuf.append(X[i]) yshuf.append(y[i]) Xshuf = np.array(Xshuf) yshuf = np.array(yshuf) #TODO: handle the bimodal case. return Xshuf,yshuf
