def _update_hist(self, new_input):
range_ext = cp.around(new_input.min() - self.bin_size / 2, 1), \
cp.around(new_input.max() + self.bin_size / 2, 1)
bins_array = cp.arange(range_ext[0], range_ext[1] + self.bin_size,
self.bin_size)
weights, bins = cp.histogram(new_input, bins_array)
if self._empty:
self.weights, self.bins = weights, bins[:-1]
self._empty = False
else: # update the hist
self.weights, self.bins = concat_hists(self.weights, self.bins,
weights, bins[:-1],
self.bin_size, self._rd)
def _fft_convolve(a1, a2, mode):
if a1.size < a2.size:
a1, a2 = a2, a1
if a1.dtype.kind == 'c' or a2.dtype.kind == 'c':
fft, ifft = cupy.fft.fft, cupy.fft.ifft
else:
fft, ifft = cupy.fft.rfft, cupy.fft.irfft
dtype = cupy.result_type(a1, a2)
n1, n2 = a1.size, a2.size
out_size = n1 + n2 - 1
fa1 = fft(a1, out_size)
fa2 = fft(a2, out_size)
out = ifft(fa1 * fa2, out_size)
if mode == 'full':
start, end = None, None
elif mode == 'same':
start = (n2 - 1) // 2
end = start + n1
elif mode == 'valid':
start, end = n2 - 1, n1
else:
raise ValueError(
'acceptable mode flags are `valid`, `same`, or `full`.')
out = out[start:end]
if dtype.kind in 'iu':
out = cupy.around(out)
return out.astype(dtype, copy=False)
def initialize(m, n, d):
W = cp.random.normal(size=(m, d))
y = cp.random.normal(size=n)
a = 2*cp.around(cp.random.uniform(size=m))-cp.ones(m)
x_tmp = cp.random.uniform(size=(n, d))
x = cp.divide(x_tmp, cp.linalg.norm(x_tmp))
return x, W, a, y
def eval_emo_lex(derived_emo_lex, emo_lex, trans, induct_emos_file, induct_emos_eval_file, emotion):
print("Number of derived emotion ratings:", len(derived_emo_lex))
derived_emos = []
real_emos = []
words = []
trans = {word_src: tgt_words for word_src, tgt_words in trans}
for word, emo in derived_emo_lex.items():
translations = ",".join([t[0] for t in trans[word]])
induct_emos_file.write(f"{word}\t{translations}\t{emotion}\t{emo}\n")
real_emo = emo_lex.get(word, None)
if real_emo:
induct_emos_eval_file.write(f"{word}\t{translations}\t{emotion}\t{emo}\t{real_emo}\n")
derived_emos.append(emo)
real_emos.append(real_emo)
words.append(word)
print("Coverage in test set:", len(derived_emos) / len(derived_emo_lex))
derived_emos = np.array(derived_emos, dtype=float)
real_emos = np.array(real_emos, dtype=float)
corr_coeff = np.corrcoef(derived_emos, real_emos, rowvar=False)
top_words = np.argsort(-derived_emos)[:10]
print(derived_emos[top_words])
top_words = [words[int(idx)] for idx in top_words]
print(top_words)
corr_coeff = np.around(corr_coeff[0, 1], 3)
print("Correlation:", corr_coeff)
return [corr_coeff, len(derived_emo_lex), derived_emos.shape[0]]
def FSITM(HDR, LDR, alpha=None):
NumPixels = LDR.size
if alpha is None:
r = cp.floor(NumPixels / (2.**18))
if r > 1.:
alpha = 1. - (1. / r)
else:
alpha = 0.
minNonzero = cp.min(HDR[HDR > 0])
LogH = cp.log(cp.maximum(HDR, minNonzero))
# float is needed for further calculation
LogH = cp.around((LogH - LogH.min()) * 255. /
(LogH.max() - LogH.min())).astype(cp.float)
if alpha > 0.:
PhaseHDR_CH = phasecong100(HDR, 2, 2, 8, 8)
PhaseLDR_CH8 = phasecong100(LDR, 2, 2, 8, 8)
else: # so, if image size is smaller than 512x512?
PhaseHDR_CH = 0
PhaseLDR_CH8 = 0
PhaseLogH = phasecong100(LogH, 2, 2, 2, 2)
PhaseH = alpha * PhaseHDR_CH + (1 - alpha) * PhaseLogH
PhaseLDR_CH2 = phasecong100(LDR, 2, 2, 2, 2)
PhaseL = alpha * PhaseLDR_CH8 + (1 - alpha) * PhaseLDR_CH2
Q = cp.sum(
cp.logical_or(cp.logical_and(PhaseL <= 0, PhaseH <= 0),
cp.logical_and(PhaseL > 0, PhaseH > 0))) / NumPixels
return Q
def gpd(points, fpoints=fpoints, cell=cell, radii=radii):
points = cp.asarray(points)
diff = cp.expand_dims(points, axis=1)-cp.expand_dims(fpoints, axis=0)
diff = (diff - cp.around(diff)).reshape(-1,3)
diff = cp.dot(cell, diff.T).T
diff = cp.linalg.norm(diff, axis=1).reshape(-1,fpoints.shape[0])-radii.T
return cp.min(diff, axis=1)
def test_probe_support(self):
"""Finite probe support penalty function is within expected bounds."""
penalty = tike.ptycho.probe.finite_probe_support(
probe=cp.zeros((101, 101)), # must be odd shaped for min to be 0
radius=0.5 * 0.4,
degree=1.0, # must have degree >= 1 for upper bound to be p
p=2.345,
)
try:
import tifffile
os.makedirs(resultdir, exist_ok=True)
tifffile.imsave(os.path.join(resultdir, 'penalty.tiff'),
penalty.astype('float32').get())
except ImportError:
pass
assert cp.around(cp.min(penalty), 3) == 0.000
assert cp.around(cp.max(penalty), 3) == 2.345
def draw_beta(min_value, max_value, mean_value, n_values, round=False):
""" draw `n_values` values between `min_value` and `max_value` having
`mean_value` as (asymptotical) average"""
a, b = get_alpha_beta(min_value, max_value, mean_value)
durations = cp.random.beta(
a, b, n_values) * (max_value - min_value) + min_value
if round:
durations = cp.around(durations)
return durations.reshape(-1, 1).astype(cp.float16)
def counts(y):
_, y_indices = cp.unique(y, return_inverse=True)
class_counts = cp.bincount(y_indices)
total = cp.sum(class_counts)
percent_counts = []
for count in (class_counts):
percent_counts.append(
cp.around(float(count) / total.item(), decimals=2).item())
return percent_counts
def around(inp) -> 'Tensor':
_check_tensors(inp)
engine = _get_engine(inp)
return _create_tensor(
inp,
data=engine.around(inp.data),
func=wrapped_partial(around_backward, inp=inp)
)
def regression(self, test):
mean = np.zeros(test.shape[0])
for x in range(test.shape[0]): # for each test example
mean[x] = np.sum(self.n_weights[x] * self.n_labels[x])
if self.weighted: # divide by sum of the weights
mean[x] = mean[x] / np.sum(self.n_weights[x])
else: # divide by k
mean[x] = mean[x] / self.k
# print("Prediction for Test=" + str(x) + " is " + str(mean[x]))
mean = np.around(mean, 0) # round for picking a class
return mean
def zero_one(y, target, cuda=False):
if cuda:
res = (target != cp.around(y)).mean()
cp.cuda.Stream.null.synchronize()
return res
else:
y_one_hot = np.zeros(y.shape)
i = 0
for col in y:
y_one_hot[i, np.argmax(col)] = 1
i += 1
return sklearn.metrics.zero_one_loss(np.around(y_one_hot), target)
def test_silhouette_samples_batched(metric, chunk_divider, labeled_clusters):
X, labels = labeled_clusters
cuml_scores = cu_silhouette_samples(X, labels, metric=metric,
chunksize=int(X.shape[0] /
chunk_divider))
sk_scores = sk_silhouette_samples(X, labels, metric=metric)
cu_trunc = cp.around(cuml_scores, decimals=3)
sk_trunc = cp.around(sk_scores, decimals=3)
diff = cp.absolute(cu_trunc - sk_trunc) > 0
over_diff = cp.all(diff)
# 0.5% elements allowed to be different
if len(over_diff.shape) > 0:
assert over_diff.shape[0] <= 0.005 * X.shape[0]
# different elements should not differ more than 1e-1
tolerance_diff = cp.absolute(cu_trunc[diff] - sk_trunc[diff]) > 1e-1
diff_change = cp.all(tolerance_diff)
if len(diff_change.shape) > 0:
assert False
def __write_to_txtfileIDCT(self, y8x8, cb8x8, cr8x8):
"""writing the values 8x8 blocks of each IDCT-pic (y, cb, cr) to a textfile"""
y8x8 = np.around(y8x8, 3)
cb8x8 = np.around(cb8x8, 3)
cr8x8 = np.around(cr8x8, 3)
if args.cupy:
y8x8 = np.asnumpy(y8x8)
cb8x8 = np.asnumpy(cb8x8)
cr8x8 = np.asnumpy(cr8x8)
with open(os.path.join(inputpath, args.output, "debug", infilename + "_IDCT_y.txt"), "w") as IDCT_y_txt:
with open(os.path.join(inputpath, args.output, "debug", infilename + "_IDCT_cb.txt"), "w") as IDCT_cb_txt:
with open(os.path.join(inputpath, args.output, "debug", infilename + "_IDCT_cr.txt"), "w") as IDCT_cr_txt:
for i, block in enumerate(y8x8):
IDCT_y_txt.write("Block: " + str(i+1) + "\n")
IDCT_cb_txt.write("Block: " + str(i+1) + "\n")
IDCT_cr_txt.write("Block: " + str(i+1) + "\n")
for j in range(64):
IDCT_y_txt.write(str(j+1) + ": " + str(y8x8[i][j]) + "; ")
IDCT_cb_txt.write(str(j+1) + ": " + str(cb8x8[i][j]) + "; ")
IDCT_cr_txt.write(str(j+1) + ": " + str(cr8x8[i][j]) + "; ")
IDCT_y_txt.write("\n")
IDCT_cb_txt.write("\n")
IDCT_cr_txt.write("\n")
def quantize_real(x,
target_mean=0,
target_std=32 / (2 * xp.sqrt(2 * xp.log(2))),
num_bits=8,
data_mean=None,
data_std=None,
stats_calc_num_samples=10000):
"""
Quantize real voltage data to integers with specified number of bits
and target statistics.
Parameters
----------
x : array
Array of voltages
target_mean : float, optional
Target mean for voltages
target_std : float, optional
Target standard deviation for voltages
num_bits : int, optional
Number of bits to quantize to. Quantized voltages will span -2**(num_bits - 1)
to 2**(num_bits - 1) - 1, inclusive.
data_mean : float, optional
Mean of input voltages, if already known
data_std : float, optional
Standard deviation of input voltages, if already known. If None, estimates mean and
standard deviation automatically.
stats_calc_num_samples : int, optional
Maximum number of samples for use in estimating noise statistics
Returns
-------
q_voltages : array
Array of quantized voltages
"""
if data_std is None:
data_mean, data_std = data_stream.estimate_stats(
x, stats_calc_num_samples)
if data_std == 0:
factor = 0
else:
factor = target_std / data_std
q_voltages = xp.around(factor * (x - data_mean) + target_mean)
q_voltages = xp.clip(q_voltages, -2**(num_bits - 1), 2**(num_bits - 1) - 1)
q_voltages = q_voltages.astype(int)
return q_voltages
def train_cv_split(x_array, y_array, cv_pct):
'''
Inputs - Dataset in the form of np array, percentage of dataset you would like to select as cross validation set \n
There are two "dataset" arguments, data and data2. This is in case you have separate arrays for x and y. They will be sampled with the same indices.
Outputs - Training set, cross validation set, output format is np array
'''
idx = cp.random.randint(x_array.shape[0], size= int(cp.around(x_array.shape[0]*cv_pct)) )
mask = cp.ones(x_array.shape[0],dtype=bool)
mask[idx] = 0
x_array_test = x_array[idx, :]
x_array_train = x_array[mask, :]
y_array_test = y_array[idx, :]
y_array_train = y_array[mask, :]
return x_array_train, x_array_test, y_array_train, y_array_test
def fit(self,
X,
Y,
learning_rate=10e-7,
reg=0,
epochs=120000,
show_fig=False):
X, Y = shuffle(X, Y)
Xvalid, Yvalid = X[-1000:], Y[-1000:]
X, Y = X[:-1000], Y[:-1000]
N, D = X.shape
self.w = cp.random.randn(D) / cp.sqrt(D)
self.b = 0
costs = []
best_validation_error = 1
for i in range(epochs):
pY = self.forward(X)
# gradient descent step
self.w -= learning_rate * (X.T.dot(pY - Y) + reg * self.w)
self.b -= learning_rate * ((pY - Y).sum() + reg * self.b)
if i % 20 == 0:
pYvalid = self.forward(Xvalid)
c = sigmoid_cost(Yvalid, pYvalid)
costs.append(c)
e = error_rate(
Yvalid, cp.around(pYvalid)
) # cp.round just means threshold of 0.5 for classification
print("i: ", i, " cost: ", c, " error: ", e)
if e < best_validation_error:
best_validation_error = e
print("best validation error: ", best_validation_error)
if show_fig:
plt.plot(costs)
plt.show()
def _fft_convolve(a1, a2, mode):
offset = 0
if a1.size < a2.size:
a1, a2 = a2, a1
offset = 1 - a2.size % 2
# if either of them is complex, the dtype after multiplication will also be
if a1.dtype.kind == 'c' or a2.dtype.kind == 'c':
fft, ifft = cupy.fft.fft, cupy.fft.ifft
else:
fft, ifft = cupy.fft.rfft, cupy.fft.irfft
dtype = cupy.result_type(a1, a2)
n1, n2 = a1.size, a2.size
out_size = cupyx.scipy.fft.next_fast_len(n1 + n2 - 1)
fa1 = fft(a1, out_size)
fa2 = fft(a2, out_size)
out = ifft(fa1 * fa2, out_size)
if mode == 'full':
start, end = 0, n1 + n2 - 1
elif mode == 'same':
start = (n2 - 1) // 2 + offset
end = start + n1
elif mode == 'valid':
start, end = n2 - 1, n1
else:
raise ValueError(
'acceptable mode flags are `valid`, `same`, or `full`.')
out = out[start:end]
if dtype.kind in 'iu':
out = cupy.around(out)
return out.astype(dtype, copy=False)
def test_mark_boundaries_subpixel():
# fmt: off
labels = cp.array(
[[0, 0, 0, 0], [0, 0, 5, 0], [0, 1, 5, 0], [0, 0, 5, 0], [0, 0, 0, 0]],
dtype=np.uint8)
np.random.seed(0)
# fmt: on
# Note: use np.random to have same seed as NumPy
# Note: use np.round until cp.around is fixed upstream
image = cp.asarray(np.round(np.random.rand(*labels.shape), 2))
marked = mark_boundaries(image, labels, color=white, mode='subpixel')
marked_proj = cp.asarray(cp.around(cp.mean(marked, axis=-1), 2))
# fmt: off
ref_result = cp.array([
[0.55, 0.63, 0.72, 0.69, 0.6, 0.55, 0.54], # noqa
[0.45, 0.58, 0.72, 1., 1., 1., 0.69], # noqa
[0.42, 0.54, 0.65, 1., 0.44, 1., 0.89], # noqa
[0.69, 1., 1., 1., 0.69, 1., 0.83], # noqa
[0.96, 1., 0.38, 1., 0.79, 1., 0.53], # noqa
[0.89, 1., 1., 1., 0.38, 1., 0.16], # noqa
[0.57, 0.78, 0.93, 1., 0.07, 1., 0.09], # noqa
[0.2, 0.52, 0.92, 1., 1., 1., 0.54], # noqa
[0.02, 0.35, 0.83, 0.9, 0.78, 0.81, 0.87]
]) # noqa
# fmt: on
# TODO: get fully equivalent interpolation/boundary as skimage
# I think this requires fixing mode='reflect' upstream in SciPy
if False:
assert_allclose(marked_proj, ref_result, atol=0.01)
else:
# Note: grlee77: only test locations of ones, due to different default
# interpolation settings in CuPy version of mark
# boundaries
assert_allclose(marked_proj == 1, ref_result == 1, atol=0.01)
def phase_cross_correlation(reference_image,
moving_image,
*,
upsample_factor=1,
space="real",
return_error=True,
reference_mask=None,
moving_mask=None,
overlap_ratio=0.3):
"""Efficient subpixel image translation registration by cross-correlation.
This code gives the same precision as the FFT upsampled cross-correlation
in a fraction of the computation time and with reduced memory requirements.
It obtains an initial estimate of the cross-correlation peak by an FFT and
then refines the shift estimation by upsampling the DFT only in a small
neighborhood of that estimate by means of a matrix-multiply DFT.
Parameters
----------
reference_image : array
Reference image.
moving_image : array
Image to register. Must be same dimensionality as
``reference_image``.
upsample_factor : int, optional
Upsampling factor. Images will be registered to within
``1 / upsample_factor`` of a pixel. For example
``upsample_factor == 20`` means the images will be registered
within 1/20th of a pixel. Default is 1 (no upsampling).
Not used if any of ``reference_mask`` or ``moving_mask`` is not None.
space : string, one of "real" or "fourier", optional
Defines how the algorithm interprets input data. "real" means
data will be FFT'd to compute the correlation, while "fourier"
data will bypass FFT of input data. Case insensitive. Not
used if any of ``reference_mask`` or ``moving_mask`` is not
None.
return_error : bool, optional
Returns error and phase difference if on, otherwise only
shifts are returned. Has noeffect if any of ``reference_mask`` or
``moving_mask`` is not None. In this case only shifts is returned.
reference_mask : ndarray
Boolean mask for ``reference_image``. The mask should evaluate
to ``True`` (or 1) on valid pixels. ``reference_mask`` should
have the same shape as ``reference_image``.
moving_mask : ndarray or None, optional
Boolean mask for ``moving_image``. The mask should evaluate to ``True``
(or 1) on valid pixels. ``moving_mask`` should have the same shape
as ``moving_image``. If ``None``, ``reference_mask`` will be used.
overlap_ratio : float, optional
Minimum allowed overlap ratio between images. The correlation for
translations corresponding with an overlap ratio lower than this
threshold will be ignored. A lower `overlap_ratio` leads to smaller
maximum translation, while a higher `overlap_ratio` leads to greater
robustness against spurious matches due to small overlap between
masked images. Used only if one of ``reference_mask`` or
``moving_mask`` is None.
Returns
-------
shifts : ndarray
Shift vector (in pixels) required to register ``moving_image``
with ``reference_image``. Axis ordering is consistent with
numpy (e.g. Z, Y, X)
error : float
Translation invariant normalized RMS error between
``reference_image`` and ``moving_image``.
phasediff : float
Global phase difference between the two images (should be
zero if images are non-negative).
References
----------
.. [1] Manuel Guizar-Sicairos, Samuel T. Thurman, and James R. Fienup,
"Efficient subpixel image registration algorithms,"
Optics Letters 33, 156-158 (2008). :DOI:`10.1364/OL.33.000156`
.. [2] James R. Fienup, "Invariant error metrics for image reconstruction"
Optics Letters 36, 8352-8357 (1997). :DOI:`10.1364/AO.36.008352`
.. [3] Dirk Padfield. Masked Object Registration in the Fourier Domain.
IEEE Transactions on Image Processing, vol. 21(5),
pp. 2706-2718 (2012). :DOI:`10.1109/TIP.2011.2181402`
.. [4] D. Padfield. "Masked FFT registration". In Proc. Computer Vision and
Pattern Recognition, pp. 2918-2925 (2010).
:DOI:`10.1109/CVPR.2010.5540032`
"""
if (reference_mask is not None) or (moving_mask is not None):
return _masked_phase_cross_correlation(reference_image, moving_image,
reference_mask, moving_mask,
overlap_ratio)
# images must be the same shape
if reference_image.shape != moving_image.shape:
raise ValueError("images must be same shape")
# assume complex data is already in Fourier space
if space.lower() == 'fourier':
src_freq = reference_image
target_freq = moving_image
# real data needs to be fft'd.
elif space.lower() == 'real':
src_freq = fft.fftn(reference_image)
target_freq = fft.fftn(moving_image)
else:
raise ValueError('space argument must be "real" of "fourier"')
# Whole-pixel shift - Compute cross-correlation by an IFFT
shape = src_freq.shape
image_product = src_freq * target_freq.conj()
cross_correlation = fft.ifftn(image_product)
# Locate maximum
maxima = cp.unravel_index(cp.argmax(cp.abs(cross_correlation)),
cross_correlation.shape)
midpoints = cp.asarray([np.fix(axis_size / 2) for axis_size in shape])
float_dtype = image_product.real.dtype
shifts = cp.stack([m.astype(float_dtype, copy=False) for m in maxima])
shifts[shifts > midpoints] -= cp.asarray(shape)[shifts > midpoints]
if upsample_factor == 1:
if return_error:
sabs = cp.abs(src_freq)
sabs *= sabs
tabs = cp.abs(target_freq)
tabs *= tabs
src_amp = np.sum(sabs) / src_freq.size
target_amp = np.sum(tabs) / target_freq.size
CCmax = cross_correlation[maxima]
# If upsampling > 1, then refine estimate with matrix multiply DFT
else:
# Initial shift estimate in upsampled grid
shifts = cp.around(shifts * upsample_factor) / upsample_factor
upsampled_region_size = math.ceil(upsample_factor * 1.5)
# Center of output array at dftshift + 1
dftshift = np.fix(upsampled_region_size / 2.0)
upsample_factor = float(upsample_factor)
# Matrix multiply DFT around the current shift estimate
sample_region_offset = dftshift - shifts * upsample_factor
cross_correlation = _upsampled_dft(image_product.conj(),
upsampled_region_size,
upsample_factor,
sample_region_offset).conj()
# Locate maximum and map back to original pixel grid
maxima = cp.unravel_index(cp.argmax(cp.abs(cross_correlation)),
cross_correlation.shape)
CCmax = cross_correlation[maxima]
maxima = (cp.stack([m.astype(float_dtype, copy=False)
for m in maxima]) - dftshift)
shifts = shifts + maxima / upsample_factor
if return_error:
src_amp = cp.abs(src_freq)
src_amp *= src_amp
src_amp = cp.sum(src_amp)
target_amp = cp.abs(target_freq)
target_amp *= target_amp
target_amp = cp.sum(target_amp)
# If its only one row or column the shift along that dimension has no
# effect. We set to zero.
for dim in range(src_freq.ndim):
if shape[dim] == 1:
shifts[dim] = 0
if return_error:
# Redirect user to masked_phase_cross_correlation if NaNs are observed
if cp.isnan(CCmax) or cp.isnan(src_amp) or cp.isnan(target_amp):
raise ValueError(
"NaN values found, please remove NaNs from your "
"input data or use the `reference_mask`/`moving_mask` "
"keywords, eg: "
"phase_cross_correlation(reference_image, moving_image, "
"reference_mask=~np.isnan(reference_image), "
"moving_mask=~np.isnan(moving_image))")
return shifts, _compute_error(CCmax, src_amp, target_amp),\
_compute_phasediff(CCmax)
else:
return shifts
# fY11=reduce_y(flag=2)
z=r+jay*rx
yy=1/z
cp.fill_diagonal(yyfull, yy)
Y_dummy=cp.matmul(yyfull,c_line.T)
Y=cp.matmul(c_line,Y_dummy)+Y_2
cp.fill_diagonal(Y,Y.diagonal()+yl+Gb+jay*Bb)
Y[g_bus[:,None]-1,g_bus-1]=Y[g_bus[:,None]-1,g_bus-1]+permmod[:]
posfrecV1=-cp.linalg.solve(Y,Y_c)
temp=cp.matmul(Y_b,posfrecV1)
posfY11=Y_a+temp.T
#print(posfY11)
#print("Finish reduced Y, start simulation!")
# Start of simulation
t_step=cp.around((s1[1:]-s1[:-1])/s7[:-1])
t_width=(s1[1:]-s1[:-1])/t_step[:]
sim_k=int(t_step.sum())
sim_k=sim_k+1
#print(sim_k)
mac_ang,mac_spd,dmac_ang,dmac_spd,pelect=cp.zeros((5,sim_k,ngen),dtype=cp.float64)
eprime=cp.zeros((sim_k,ngen),dtype=cp.complex128)
theta=cp.radians(b_ang)
bus_volt=V*cp.exp(jay*theta)
mva=basmva/g_m
tst1=bus_int[g_bus-1].astype(cp.int)
eterm=V[tst1-1] # terminal bus voltage
pelect[0]=b_pg[tst1-1] # BUS_pg
qelect=b_qg[tst1-1] # BUS_qg
#compute the initial values for generator dynamics
def predict(self, X):
pY = self.forward(X)
return cp.around(pY) # binary classification with threshold of 0.5
def _fft_convolve(a1, a2, mode):
offset = 0
if a1.size < a2.size:
a1, a2 = a2, a1
offset = 1 - a2.size % 2
# if either of them is complex, the dtype after multiplication will also be
if a1.dtype.kind == 'c' or a2.dtype.kind == 'c':
fft, ifft = cupy.fft.fft, cupy.fft.ifft
is_c2c = True
else:
fft, ifft = cupy.fft.rfft, cupy.fft.irfft
is_c2c = False
# hack to work around NumPy/CuPy FFT dtype incompatibility:
# CuPy internally converts fp16 to fp32 before doing FFT (whereas Numpy
# converts both fp16 and fp32 to fp64), so here we do the cast early and
# explicitly, and make sure a correct cuFFT plan can be generated. After
# the fft-ifft round trip, we cast the output dtype to the correct one.
out_dtype = cupy.result_type(a1, a2)
dtype = _output_dtype(out_dtype, 'C2C' if is_c2c else 'R2C')
a1 = a1.astype(dtype, copy=False)
a2 = a2.astype(dtype, copy=False)
n1, n2 = a1.size, a2.size
out_size = cupyx.scipy.fft.next_fast_len(n1 + n2 - 1)
# skip calling get_fft_plan() as we know the args exactly
if is_c2c:
fft_t = cufft.CUFFT_C2C if dtype == cupy.complex64 else cufft.CUFFT_Z2Z
fft_plan = cufft.Plan1d(out_size, fft_t, 1)
ifft_plan = fft_plan
else:
fft_t = cufft.CUFFT_R2C if dtype == cupy.float32 else cufft.CUFFT_D2Z
fft_plan = cufft.Plan1d(out_size, fft_t, 1)
# this is a no-op context manager
# TODO(leofang): use contextlib.nullcontext() for PY37+?
ifft_plan = contextlib.suppress()
with fft_plan:
fa1 = fft(a1, out_size)
fa2 = fft(a2, out_size)
with ifft_plan:
out = ifft(fa1 * fa2, out_size)
if mode == 'full':
start, end = 0, n1 + n2 - 1
elif mode == 'same':
start = (n2 - 1) // 2 + offset
end = start + n1
elif mode == 'valid':
start, end = n2 - 1, n1
else:
raise ValueError(
'acceptable mode flags are `valid`, `same`, or `full`.')
out = out[start:end]
if out.dtype.kind in 'iu':
out = cupy.around(out)
return out.astype(out_dtype, copy=False)
def equalize_adapthist(image, kernel_size=None, clip_limit=0.01, nbins=256):
"""Contrast Limited Adaptive Histogram Equalization (CLAHE).
An algorithm for local contrast enhancement, that uses histograms computed
over different tile regions of the image. Local details can therefore be
enhanced even in regions that are darker or lighter than most of the image.
Parameters
----------
image : (N1, ...,NN[, C]) ndarray
Input image.
kernel_size: int or array_like, optional
Defines the shape of contextual regions used in the algorithm. If
iterable is passed, it must have the same number of elements as
``image.ndim`` (without color channel). If integer, it is broadcasted
to each `image` dimension. By default, ``kernel_size`` is 1/8 of
``image`` height by 1/8 of its width.
clip_limit : float, optional
Clipping limit, normalized between 0 and 1 (higher values give more
contrast).
nbins : int, optional
Number of gray bins for histogram ("data range").
Returns
-------
out : (N1, ...,NN[, C]) ndarray
Equalized image with float64 dtype.
See Also
--------
equalize_hist, rescale_intensity
Notes
-----
* For color images, the following steps are performed:
- The image is converted to HSV color space
- The CLAHE algorithm is run on the V (Value) channel
- The image is converted back to RGB space and returned
* For RGBA images, the original alpha channel is removed.
.. versionchanged:: 0.17
The values returned by this function are slightly shifted upwards
because of an internal change in rounding behavior.
References
----------
.. [1] http://tog.acm.org/resources/GraphicsGems/
.. [2] https://en.wikipedia.org/wiki/CLAHE#CLAHE
"""
image = img_as_uint(image)
image = cp.around(
rescale_intensity(image, out_range=(0, NR_OF_GRAY - 1))
).astype(cp.uint16)
if kernel_size is None:
kernel_size = tuple(
[image.shape[dim] // 8 for dim in range(image.ndim)]
)
elif isinstance(kernel_size, numbers.Number):
kernel_size = (kernel_size,) * image.ndim
elif len(kernel_size) != image.ndim:
ValueError("Incorrect value of `kernel_size`: {}".format(kernel_size))
kernel_size = [int(k) for k in kernel_size]
image = _clahe(image, kernel_size, clip_limit, nbins)
image = img_as_float(image)
return rescale_intensity(image)
def cross_correlate_masked(arr1,
arr2,
m1,
m2,
mode="full",
axes=(-2, -1),
overlap_ratio=0.3):
"""
Masked normalized cross-correlation between arrays.
Parameters
----------
arr1 : ndarray
First array.
arr2 : ndarray
Seconds array. The dimensions of `arr2` along axes that are not
transformed should be equal to that of `arr1`.
m1 : ndarray
Mask of `arr1`. The mask should evaluate to `True`
(or 1) on valid pixels. `m1` should have the same shape as `arr1`.
m2 : ndarray
Mask of `arr2`. The mask should evaluate to `True`
(or 1) on valid pixels. `m2` should have the same shape as `arr2`.
mode : {'full', 'same'}, optional
'full':
This returns the convolution at each point of overlap. At
the end-points of the convolution, the signals do not overlap
completely, and boundary effects may be seen.
'same':
The output is the same size as `arr1`, centered with respect
to the `‘full’` output. Boundary effects are less prominent.
axes : tuple of ints, optional
Axes along which to compute the cross-correlation.
overlap_ratio : float, optional
Minimum allowed overlap ratio between images. The correlation for
translations corresponding with an overlap ratio lower than this
threshold will be ignored. A lower `overlap_ratio` leads to smaller
maximum translation, while a higher `overlap_ratio` leads to greater
robustness against spurious matches due to small overlap between
masked images.
Returns
-------
out : ndarray
Masked normalized cross-correlation.
Raises
------
ValueError : if correlation `mode` is not valid, or array dimensions along
non-transformation axes are not equal.
References
----------
.. [1] Dirk Padfield. Masked Object Registration in the Fourier Domain.
IEEE Transactions on Image Processing, vol. 21(5),
pp. 2706-2718 (2012). :DOI:`10.1109/TIP.2011.2181402`
.. [2] D. Padfield. "Masked FFT registration". In Proc. Computer Vision and
Pattern Recognition, pp. 2918-2925 (2010).
:DOI:`10.1109/CVPR.2010.5540032`
"""
if mode not in {"full", "same"}:
raise ValueError("Correlation mode {} is not valid.".format(mode))
if arr1.dtype.kind == "c" or arr2.dtype.kind == "c":
raise ValueError("complex-valued arr1, arr2 are not supported")
fixed_image = cp.asarray(arr1, dtype=np.float)
fixed_mask = cp.asarray(m1, dtype=np.bool)
moving_image = cp.asarray(arr2, dtype=np.float)
moving_mask = cp.asarray(m2, dtype=np.bool)
eps = np.finfo(np.float).eps
# Array dimensions along non-transformation axes should be equal.
all_axes = set(range(fixed_image.ndim))
for axis in all_axes - set(axes):
if fixed_image.shape[axis] != moving_image.shape[axis]:
raise ValueError(
"Array shapes along non-transformation axes should be "
"equal, but dimensions along axis {a} are not".format(a=axis))
# Determine final size along transformation axes
# Note that it might be faster to compute Fourier transform in a slightly
# larger shape (`fast_shape`). Then, after all fourier transforms are done,
# we slice back to`final_shape` using `final_slice`.
final_shape = list(arr1.shape)
for axis in axes:
final_shape[axis] = (fixed_image.shape[axis] +
moving_image.shape[axis] - 1)
final_shape = tuple(final_shape)
final_slice = tuple([slice(0, int(sz)) for sz in final_shape])
# Extent transform axes to the next fast length (i.e. multiple of 3, 5, or
# 7)
fast_shape = tuple([next_fast_len(final_shape[ax]) for ax in axes])
# We use numpy.fft or the new scipy.fft because they allow leaving the
# transform axes unchanged which was not possible with scipy.fftpack's
# fftn/ifftn in older versions of SciPy.
# E.g. arr shape (2, 3, 7), transform along axes (0, 1) with shape (4, 4)
# results in arr_fft shape (4, 4, 7)
fft = partial(fftmodule.fftn, s=fast_shape, axes=axes)
ifft = partial(fftmodule.ifftn, s=fast_shape, axes=axes)
fixed_image[cp.logical_not(fixed_mask)] = 0.0
moving_image[cp.logical_not(moving_mask)] = 0.0
# N-dimensional analog to rotation by 180deg is flip over all relevant axes.
# See [1] for discussion.
rotated_moving_image = _flip(moving_image, axes=axes)
rotated_moving_mask = _flip(moving_mask, axes=axes)
fixed_fft = fft(fixed_image)
rotated_moving_fft = fft(rotated_moving_image)
fixed_mask_fft = fft(fixed_mask)
rotated_moving_mask_fft = fft(rotated_moving_mask)
# Calculate overlap of masks at every point in the convolution.
# Locations with high overlap should not be taken into account.
number_overlap_masked_px = cp.real(
ifft(rotated_moving_mask_fft * fixed_mask_fft))
number_overlap_masked_px[:] = cp.around(number_overlap_masked_px)
number_overlap_masked_px[:] = cp.fmax(number_overlap_masked_px, eps)
masked_correlated_fixed_fft = ifft(rotated_moving_mask_fft * fixed_fft)
masked_correlated_rotated_moving_fft = ifft(fixed_mask_fft *
rotated_moving_fft)
numerator = ifft(rotated_moving_fft * fixed_fft)
numerator -= (masked_correlated_fixed_fft *
masked_correlated_rotated_moving_fft /
number_overlap_masked_px)
fixed_squared_fft = fft(cp.square(fixed_image))
fixed_denom = ifft(rotated_moving_mask_fft * fixed_squared_fft)
fixed_denom -= (cp.square(masked_correlated_fixed_fft) /
number_overlap_masked_px)
fixed_denom[:] = cp.fmax(fixed_denom, 0.0)
rotated_moving_squared_fft = fft(cp.square(rotated_moving_image))
moving_denom = ifft(fixed_mask_fft * rotated_moving_squared_fft)
moving_denom -= (cp.square(masked_correlated_rotated_moving_fft) /
number_overlap_masked_px)
moving_denom[:] = cp.fmax(moving_denom, 0.0)
denom = cp.sqrt(fixed_denom * moving_denom)
# Slice back to expected convolution shape.
numerator = numerator[final_slice]
denom = denom[final_slice]
number_overlap_masked_px = number_overlap_masked_px[final_slice]
if mode == "same":
_centering = partial(_centered, newshape=fixed_image.shape, axes=axes)
denom = _centering(denom)
numerator = _centering(numerator)
number_overlap_masked_px = _centering(number_overlap_masked_px)
# Pixels where `denom` is very small will introduce large
# numbers after division. To get around this problem,
# we zero-out problematic pixels.
tol = 1e3 * eps * cp.max(cp.abs(denom), axis=axes, keepdims=True)
nonzero_indices = denom > tol
# TODO: grlee77: Added a cast to real here.
# probably it should be real earlier?
numerator = numerator.real
denom = denom.real
out = cp.zeros_like(denom)
out[nonzero_indices] = numerator[nonzero_indices] / denom[nonzero_indices]
cp.clip(out, a_min=-1, a_max=1, out=out)
# Apply overlap ratio threshold
number_px_threshold = overlap_ratio * cp.max(
number_overlap_masked_px, axis=axes, keepdims=True)
out[number_overlap_masked_px < number_px_threshold] = 0.0
return out
def convolve(
in1,
in2,
mode="full",
method="auto",
):
"""
Convolve two N-dimensional arrays.
Convolve `in1` and `in2`, with the output size determined by the
`mode` argument.
Parameters
----------
in1 : array_like
First input.
in2 : array_like
Second input. Should have the same number of dimensions as `in1`.
mode : str {'full', 'valid', 'same'}, optional
A string indicating the size of the output:
``full``
The output is the full discrete linear convolution
of the inputs. (Default)
``valid``
The output consists only of those elements that do not
rely on the zero-padding. In 'valid' mode, either `in1` or `in2`
must be at least as large as the other in every dimension.
``same``
The output is the same size as `in1`, centered
with respect to the 'full' output.
method : str {'auto', 'direct', 'fft'}, optional
A string indicating which method to use to calculate the convolution.
``direct``
The convolution is determined directly from sums, the definition of
convolution.
``fft``
The Fourier Transform is used to perform the convolution by calling
`fftconvolve`.
``auto``
Automatically chooses direct or Fourier method based on an estimate
of which is faster (default).
Returns
-------
convolve : array
An N-dimensional array containing a subset of the discrete linear
convolution of `in1` with `in2`.
See Also
--------
choose_conv_method : chooses the fastest appropriate convolution method
fftconvolve
Notes
-----
By default, `convolve` and `correlate` use ``method='auto'``, which calls
`choose_conv_method` to choose the fastest method using pre-computed
values (`choose_conv_method` can also measure real-world timing with a
keyword argument). Because `fftconvolve` relies on floating point numbers,
there are certain constraints that may force `method=direct` (more detail
in `choose_conv_method` docstring).
Examples
--------
Smooth a square pulse using a Hann window:
>>> import cusignal
>>> import cupy as cp
>>> sig = cp.repeat(cp.asarray([0., 1., 0.]), 100)
>>> win = cusignal.hann(50)
>>> filtered = cusignal.convolve(sig, win, mode='same') / cp.sum(win)
>>> import matplotlib.pyplot as plt
>>> fig, (ax_orig, ax_win, ax_filt) = plt.subplots(3, 1, sharex=True)
>>> ax_orig.plot(cp.asnumpy(sig))
>>> ax_orig.set_title('Original pulse')
>>> ax_orig.margins(0, 0.1)
>>> ax_win.plot(cp.asnumpy(win))
>>> ax_win.set_title('Filter impulse response')
>>> ax_win.margins(0, 0.1)
>>> ax_filt.plot(cp.asnumpy(filtered))
>>> ax_filt.set_title('Filtered signal')
>>> ax_filt.margins(0, 0.1)
>>> fig.tight_layout()
>>> fig.show()
"""
volume = cp.asarray(in1)
kernel = cp.asarray(in2)
if volume.ndim == kernel.ndim == 0:
return volume * kernel
elif volume.ndim != kernel.ndim:
raise ValueError("in1 and in2 should have the same dimensionality")
if _inputs_swap_needed(mode, volume.shape, kernel.shape):
# Convolution is commutative
# order doesn't have any effect on output
volume, kernel = kernel, volume
if method == "auto":
method = choose_conv_method(volume, kernel, mode=mode)
if method == "fft":
out = fftconvolve(volume, kernel, mode=mode)
result_type = cp.result_type(volume, kernel)
if result_type.kind in {"u", "i"}:
out = cp.around(out)
return out.astype(result_type)
elif method == "direct":
if volume.ndim > 1:
raise ValueError("Direct method is only implemented for 1D")
swapped_inputs = (mode != "valid") and (kernel.size > volume.size)
if swapped_inputs:
volume, kernel = kernel, volume
return _convolution_cuda._convolve(volume, kernel, True,
swapped_inputs, mode)
else:
raise ValueError("Acceptable method flags are 'auto',"
" 'direct', or 'fft'.")
def concat_hists(weights1, bins1, weights2, bins2, bin_size, rd):
min1, max1 = cp.around(bins1[0], rd), cp.around(bins1[-1], rd)
min2, max2 = cp.around(bins2[0], rd), cp.around(bins2[-1], rd)
mini, maxi = min(min1, min2), max(max1, max2)
new_bins = cp.arange(
mini, maxi + bin_size * 0.9,
bin_size) # * 0.9 to avoid unexpected random inclusion of last element
if min1 - mini != 0 and maxi - max1 != 0:
ext1 = cp.pad(weights1, (cp.int(cp.around((min1 - mini) / bin_size)),
cp.int(cp.around((maxi - max1) / bin_size))),
'constant',
constant_values=0)
elif min1 - mini != 0:
ext1 = cp.pad(weights1, (cp.int(cp.around(
(min1 - mini) / bin_size)), 0),
'constant',
constant_values=0)
elif maxi - max1 != 0:
ext1 = cp.pad(weights1, (0, cp.int(cp.around(
(maxi - max1) / bin_size))),
'constant',
constant_values=0)
else:
ext1 = weights1
if min2 - mini != 0 and maxi - max2 != 0:
ext2 = cp.pad(weights2, (cp.int(cp.around((min2 - mini) / bin_size)),
cp.int(cp.around((maxi - max2) / bin_size))),
'constant',
constant_values=0)
elif min2 - mini != 0:
ext2 = cp.pad(weights2, (cp.int(cp.around(
(min2 - mini) / bin_size)), 0),
'constant',
constant_values=0)
elif maxi - max2 != 0:
ext2 = cp.pad(weights2, (0, cp.int(cp.around(
(maxi - max2) / bin_size))),
'constant',
constant_values=0)
else:
ext2 = weights2
new_ext = ext1 + ext2
return new_ext, new_bins
def Newtons_method_feasible_init_point(f,
A,
x_0,
tol,
tol_backtracking,
x_ast=None,
p_ast=None,
maxiter=30,
gf_symbolic=None,
Hf_symbolic=None,
Sigma=None):
'''
Newton's method to numerically approximate solution of min f subject to Ax = b.
IMPORTANT: this implementation requires that initial point x_0, satisfies: Ax_0 = b
Args:
f (fun): definition of function f as lambda expression or function definition.
A (numpy ndarray): 2d numpy array of shape (m,n) defines system of constraints Ax=b.
x_0 (numpy ndarray): initial point for Newton's method. Must satisfy: Ax_0 = b
tol (float): tolerance that will halt method. Controls stopping criteria.
tol_backtracking (float): tolerance that will halt method. Controls value of line search by backtracking.
x_ast (numpy ndarray): solution of min f, now it's required that user knows the solution...
p_ast (float): value of f(x_ast), now it's required that user knows the solution...
maxiter (int): maximum number of iterations
gf_symbolic (fun): definition of gradient of f. If given, no approximation is
performed via finite differences.
Hf_symbolic (fun): definition of Hessian of f. If given, no approximation is
performed via fi
nite differences.
Returns:
x (numpy ndarray): numpy array, approximation of x_ast.
iteration (int): number of iterations.
Err_plot (numpy ndarray): numpy array of absolute error between p_ast and f(x) with x approximation
of x_ast. Useful for plotting.
x_plot (numpy ndarray): numpy array that containts in columns vector of approximations. Last column
contains x, approximation of solution. Useful for plotting.
'''
iteration = 0
x = x_0
feval = f(x)
if gf_symbolic:
gfeval = gf_symbolic(x, Sigma)
else:
gfeval = gradient_approximation(f, x)
if Hf_symbolic:
Hfeval = Hf_symbolic(x, Sigma)
else:
Hfeval = Hessian_approximation(f, x)
normgf = np.linalg.norm(gfeval)
condHf = solver.utils.condicion_cupy(Hfeval)
Err_plot_aux = np.zeros(maxiter)
Err_plot_aux[iteration] = solver.utils.compute_error(p_ast, feval)
Err = solver.utils.compute_error(x_ast, x)
if (A.ndim == 1):
p = 1
n = x.size
zero_matrix = cp.zeros(p)
first_stack = cp.column_stack((Hfeval, A.T))
second_stack = cp.row_stack(
(A.reshape(1, n).T, zero_matrix)).reshape(1, n + 1)[0]
else:
p, n = A.shape
zero_matrix = cp.zeros((p, p))
first_stack = np.column_stack((Hfeval, A.T))
second_stack = np.column_stack((A, zero_matrix))
x_plot = cp.zeros((n, maxiter))
x_plot[:, iteration] = x
system_matrix = cp.vstack((first_stack, second_stack))
zero_vector = cp.zeros(p)
rhs = cp.vstack((gfeval.reshape(n, 1), zero_vector.reshape(p, 1))).T[0]
#Newton's direction and Newton's decrement
dir_desc = cp.linalg.solve(system_matrix, -rhs)
dir_Newton = dir_desc[0:n]
dec_Newton = -gfeval.dot(dir_Newton)
w_dual_variable_estimation = dir_desc[n:(n + p)]
print(
'I\tNormgf \tNewton Decrement\tError x_ast\tError p_ast\tline search\tCondHf'
)
print('{}\t{}\t{}\t{}\t{}\t{}\t\t{}'.format(
iteration, cp.around(normgf, 4), cp.around(dec_Newton, 4),
cp.around(Err, 4), cp.around(Err_plot_aux[iteration], 4), "---",
cp.around(condHf, 4)))
stopping_criteria = dec_Newton / 2
iteration += 1
while (stopping_criteria > tol and iteration < maxiter):
der_direct = -dec_Newton
t = solver.line_search.line_search_by_backtracking(
f, dir_Newton, x, der_direct)
x = x + t * dir_Newton
feval = f(x)
if gf_symbolic:
gfeval = gf_symbolic(x, Sigma)
else:
gfeval = gradient_approximation(f, x)
if Hf_symbolic:
Hfeval = Hf_symbolic(x, Sigma)
else:
Hfeval = Hessian_approximation(f, x)
if (A.ndim == 1):
first_stack = cp.column_stack((Hfeval, A.T))
else:
first_stack = cp.column_stack((Hfeval, A.T))
system_matrix = cp.vstack((first_stack, second_stack))
rhs = cp.vstack((gfeval.reshape(n, 1), zero_vector.reshape(p, 1))).T[0]
#Newton's direction and Newton's decrement
dir_desc = cp.linalg.solve(system_matrix, -rhs)
dir_Newton = dir_desc[0:n]
dec_Newton = -gfeval.dot(dir_Newton)
w_dual_variable_estimation = dir_desc[n:(n + p)]
Err_plot_aux[iteration] = solver.utils.compute_error(p_ast, feval)
x_plot[:, iteration] = x
Err = solver.utils.compute_error(x_ast, x)
print('{}\t{}\t{}\t{}\t{}\t{}\t{}'.format(
iteration, cp.around(normgf, 4), cp.around(dec_Newton, 4),
cp.around(Err, 4), cp.around(Err_plot_aux[iteration], 4),
cp.around(t, 4), cp.around(condHf, 4)))
stopping_criteria = dec_Newton / 2
if t < tol_backtracking: #if t is less than tol_backtracking then we need to check the reason
iter_salida = iteration
iteration = maxiter - 1
iteration += 1
print('{} {}'.format("Error of x with respect to x_ast:", Err))
print('{} {}'.format("Approximate solution:", x))
cond = Err_plot_aux > np.finfo(float).eps * 10**(-2)
Err_plot = Err_plot_aux[cond]
if iteration == maxiter and t < tol_backtracking:
print(
"Backtracking value less than tol_backtracking, check approximation"
)
iteration = iter_salida
else:
if iteration == maxiter:
print("Reached maximum of iterations, check approximation")
x_plot = x_plot[:, :iteration]
return [x, iteration, Err_plot, x_plot]
def turtle():
# make a copy for later comparisons
global count_levels_combined
count_levels_combined_copy = count_levels_combined.copy()
level_target_arr = cp.random.randint(0, num_levels - 1, size=num_agents)
level_self_arr = cp.around(agent_levels_list).astype(int)
c_arr = cp.around(agent_classes_list).astype(int)
level_target_arr_plus_one = level_target_arr + 1
level_self_arr_plus_one = level_self_arr + 1
s_target_arr =level_to_salary(level_target_arr_plus_one)
s_self_arr = level_to_salary(level_self_arr_plus_one)
# log utility functions with out third/competition term
# #Calculate after tax target
length_of_range = (num_levels + 1) / len(tax_rate)
target_tax_bracket_arr = level_target_arr / length_of_range
target_tax_bracket_arr = cp.floor(target_tax_bracket_arr)
self_tax_bracket_arr = level_self_arr / length_of_range
self_tax_bracket_arr = cp.floor(self_tax_bracket_arr)
s_target_tax_rate_arr = cp.array(tax_rate)[target_tax_bracket_arr.astype(int)]
s_self_tax_rate_arr = cp.array(tax_rate)[self_tax_bracket_arr.astype(int)]
s_target_after_tax_arr = cp.multiply(s_target_arr, s_target_tax_rate_arr)
s_self_after_tax_arr = cp.multiply(s_self_arr, s_self_tax_rate_arr)
s_target_arr = s_target_after_tax_arr
s_self_arr = s_self_after_tax_arr
log_utility_payoff_target_alpha = cp.multiply(cp.asarray(agent_alpha_list), cp.log(s_target_arr))
log_utility_payoff_target_beta = cp.multiply(cp.asarray(agent_beta_list), cp.power(np.log(s_target_arr), 2))
log_utility_payoff_target_a_b = cp.subtract(log_utility_payoff_target_alpha, log_utility_payoff_target_beta)
log_utility_payoff_self_alpha = cp.multiply(cp.asarray(agent_alpha_list), cp.log(s_self_arr))
log_utility_payoff_self_beta = cp.multiply(cp.asarray(agent_beta_list), cp.power(np.log(s_self_arr), 2))
log_utility_payoff_self_a_b = cp.subtract(log_utility_payoff_self_alpha, log_utility_payoff_self_beta)
level_target_arr = cp.asnumpy(level_target_arr)
level_self_arr = cp.asnumpy(level_self_arr)
s_target_arr = cp.asnumpy(s_target_arr)
s_self_arr = cp.asnumpy(s_self_arr)
c_arr = cp.asnumpy(c_arr)
log_utility_payoff_target_a_b = cp.asnumpy(log_utility_payoff_target_a_b)
log_utility_payoff_self_a_b = cp.asnumpy(log_utility_payoff_self_a_b)
for i in range(num_agents):
#Tax
# # pick a random level as target
level_target = level_target_arr[i]
# # find levels
level_self = level_self_arr[i]
# # calculate salaries
s_target = s_target_arr[i]
s_self = s_self_arr[i]
c = c_arr[i]
# Note: agents should make decisions one by one, not all at once as
# previously programmed. Otherwise, the program will produce wrong
# results. Thus, we use count_levels_combined_copy instead of
# count_levels_combined.
num_target = count_levels_combined_copy[level_target]
num_self = count_levels_combined_copy[level_self]
alpha, beta, gamma = agent_alpha_list[i], agent_beta_list[i], agent_gamma_list[i]
# target utility & current utility
if utility_function == 'log':
payoff_target = log_utility_payoff_target_a_b[i]
payoff_target -= gamma * math.log(num_target + 1.0 / num_agents)
payoff_self = log_utility_payoff_self_a_b[i]
payoff_self -= gamma * math.log(num_self + 1.0 / num_agents)
else:
payoff_target = alpha * math.sqrt(s_target)
payoff_target -= beta * math.sqrt(s_target) ** 2
payoff_target -= gamma * math.log(num_target + 1.0 / num_agents)
payoff_self = alpha * math.sqrt(s_self)
payoff_self -= beta * math.sqrt(s_self) ** 2
payoff_self -= gamma * math.log(num_self + 1.0 / num_agents)
if payoff_target > payoff_self:
# move agent from self to target
count_levels_list[level_self, c] -= 1
count_levels_combined_copy[level_self] -= 1
count_levels_list[level_target, c] += 1
count_levels_combined_copy[level_target] += 1
agent_levels_list[i] = level_target
# calculate the least square difference of count change
loss = np.sum((count_levels_combined_copy - count_levels_combined) ** 2)
# update state variable(s)
count_levels_combined = count_levels_combined_copy
return loss
