def gen_colormap(max_alpha, max_beta, storage, NNU_X, NNU_D, nbrs, base_path, random=False, verbose=False):
base_path = "/home/brad/11.15/"
if random:
iden = "random"
else:
nnu = NNU(max_alpha, max_beta, storage)
nnu.build_index(NNU_D)
iden = nnu.name
count = np.zeros((max_alpha, max_beta))
for alpha in range(1, max_alpha + 1):
for beta in range(1, max_beta + 1):
if not random:
nnu_nbrs = nnu.index(NNU_X, alpha=alpha, beta=beta)
else:
nnu_nbrs = []
for x in X:
idxs = np.random.permutation(len(D))[: alpha * beta]
nnu_nbrs.append(idxs[np.argmax(np.abs(np.dot(D[idxs], x)))])
pct_found = len(np.where(nnu_nbrs == nbrs)[0]) / float(len(X))
count[alpha - 1, beta - 1] = pct_found
print alpha, beta, count[alpha - 1, beta - 1]
# plt.imshow(count, interpolation='nearest', vmin=0, vmax=1)
# plt.colorbar()
# fig = plt.gcf()
# fig.set_size_inches(18.5, 10.5)
# plt.savefig(base_path + iden + 'color.png')
# plt.clf()
np.save(base_path + iden + "_count", count)
D_mean = np.mean(D, axis=0)
D = D - D_mean
alphas = [1, 2, 3, 4, 5, 5, 5, 10, 10, 15, 25, 30]
betas = [1, 1, 1, 1, 1, 2, 4, 5, 10, 10, 25, 25]
storages = [Storage_Scheme.mini, Storage_Scheme.two_mini, Storage_Scheme.half]
storages = [Storage_Scheme.mini]
nnu_dists = {}
nnu_runtimes = {}
ABs = {}
#NNU
for storage in storages:
nnu = NNU(5, 5, storage)
print nnu.name
nnu.build_index(D)
nnu_dists[nnu.name] = []
nnu_runtimes[nnu.name] = []
ABs[nnu.name] = []
for alpha, beta in zip(alphas, betas):
runtime_total = 0.0
avg_abs = []
svm_xs_tr, svm_xs_t = [], []
total_matches = 0
total_samples = 0
enc_func = partial(nnu_to_bow, alpha=alpha, beta=beta, nnu=nnu, N=N)
D = D - D_mean
alphas = [1, 2, 3, 4, 5, 5, 5, 10, 10, 15, 25, 30]
betas = [1, 1, 1, 1, 1, 2, 4, 5, 10, 10, 25, 25]
storages = [Storage_Scheme.mini, Storage_Scheme.two_mini, Storage_Scheme.half]
storages = [Storage_Scheme.two_mini]
nnu_dists = {}
nnu_runtimes = {}
ABs = {}
pool = Pool(processes=4)
#NNU
for storage in storages:
nnu = NNU(30, 25, storage)
nnu.build_index(D)
print nnu.name
nnu_dists[nnu.name] = []
nnu_runtimes[nnu.name] = []
ABs[nnu.name] = []
for alpha, beta in zip(alphas, betas):
runtime_total = 0.0
avg_abs = []
svm_xs_tr, svm_xs_t = [], []
total_matches = 0
total_samples = 0
svm_xs_tr = pool.map(enc_func, X_tr)
x = x - D_mean
nbrs = np.argmax(np.abs(np.dot(D, x.T)), axis=0)
svm_X.append(util.bow(nbrs, args['D_atoms']))
clf = SVC(kernel='linear')
clf.fit(svm_X, Y)
assert False
svm_dict = {}
svm_dict['num_classes'] = len(set(Y))
svm_dict['num_clfs'] = len(clf.intercept_)
svm_dict['num_features'] = len(clf.coef_[0])
svm_dict['coefs'] = list(clf.coef_.flatten())
svm_dict['intercepts'] = list(clf.intercept_)
nnu = NNU(args['alpha'], args['beta'], storage)
nnu.build_index(D)
json_dict = {}
json_dict['nnu'] = nnu.to_dict()
json_dict['svm'] = svm_dict
if 'chunk_size' in args.keys():
chunk_size = args['chunk_size']
else:
chunk_size = None
json_str = chunk_json(json_dict, chunk_size=chunk_size)
with open(args['output_path'], 'w+') as fp:
fp.write(json_str)
