def eval_skewness(data, param):
logging.info("eval_skewness")
exp_name = "eval_skewness"
analyst = 6
param.part_size = analyst
param.ANALYST_COUNT = analyst * analyst
fov_count = 200 # fixed
skewness = [1.6, 1.8, 2.0, 2.2, 2.4]
method_list = ["grid_standard", "quad_standard", "kd_standard"]
res_cube_value = np.zeros((len(skewness), len(seed_list), len(method_list)))
for j in range(len(seed_list)):
param.seed = seed_list[j]
for i in range(len(skewness)):
param.ZIPFIAN_SKEW = skewness[i]
for k in range(len(method_list)):
if method_list[k] == "grid_standard":
tree = Grid_standard(data, param)
elif method_list[k] == "quad_standard":
tree = Quad_standard(data, param)
elif method_list[k] == "kd_standard":
tree = Kd_standard(data, param)
else:
logging.error("No such index structure!")
sys.exit(1)
tree.buildIndex()
res_cube_value[i, j, k] = optimization(tree, fov_count, seed_list[j], param)
res_value_summary = np.average(res_cube_value, axis=1)
np.savetxt(param.resdir + exp_name + dataset_identifier, res_value_summary, fmt="%.4f\t")
def eval_bandwidth(data, param):
logging.info("eval_bandwidth")
exp_name = "eval_bandwidth"
analyst = 6
param.part_size = analyst
param.ANALYST_COUNT = analyst * analyst
bandwidth = [10,15,20,25,30]
method_list = ['grid_standard', 'quad_standard', 'kd_standard']
res_cube_value = np.zeros((len(bandwidth), len(seed_list), len(method_list)))
for j in range(len(seed_list)):
param.seed = seed_list[j]
for i in range(len(bandwidth)):
for k in range(len(method_list)):
if method_list[k] == 'grid_standard':
tree = Grid_standard(data, param)
elif method_list[k] == 'quad_standard':
tree = Quad_standard(data, param)
elif method_list[k] == 'kd_standard':
tree = Kd_standard(data, param)
else:
logging.error('No such index structure!')
sys.exit(1)
tree.buildIndex()
answer = optimization(tree, bandwidth[i], seed_list[j], param)
res_cube_value[i, j, k] = answer[0]
res_value_summary = np.average(res_cube_value, axis=1)
np.savetxt(param.resdir + exp_name + dataset_identifier, res_value_summary, fmt='%.4f\t')
def eval_analyst(data, param):
logging.info("eval_analyst")
exp_name = "eval_analyst"
analyst = [4,5,6,7,8]
fov_count = 100 # fixed
method_list = ['grid_standard', 'quad_standard', 'kd_standard']
res_cube_value = np.zeros((len(analyst), len(seed_list), len(method_list)))
for j in range(len(seed_list)):
param.seed = seed_list[j]
for i in range(len(analyst)):
param.part_size = analyst[i]
param.ANALYST_COUNT = analyst[i] * analyst[i]
for k in range(len(method_list)):
if method_list[k] == 'grid_standard':
tree = Grid_standard(data, param)
elif method_list[k] == 'quad_standard':
tree = Quad_standard(data, param)
elif method_list[k] == 'kd_standard':
tree = Kd_standard(data, param)
else:
logging.error('No such index structure!')
sys.exit(1)
tree.buildIndex()
res_cube_value[i, j, k] = optimization(tree, fov_count, seed_list[j], param)
res_value_summary = np.average(res_cube_value, axis=1)
np.savetxt(param.resdir + exp_name + dataset_identifier , res_value_summary, fmt='%.4f\t')
def eval_workload(data, param):
logging.info("eval_workload")
exp_name = "eval_workload"
analyst = [4,5,6,7,8]
method_list = ['grid_standard', 'quad_standard', 'kd_standard']
res_cube_value = np.zeros((len(analyst), len(seed_list), len(method_list)))
for j in range(len(seed_list)):
param.seed = seed_list[j]
for i in range(len(analyst)):
param.part_size = analyst[i]
param.ANALYST_COUNT = analyst[i] * analyst[i]
for k in range(len(method_list)):
if method_list[k] == 'grid_standard':
tree = Grid_standard(data, param)
tree.buildIndex()
leaf_boxes = getLeafNode(tree, 1)
elif method_list[k] == 'quad_standard':
tree = Quad_standard(data, param)
tree.buildIndex()
leaf_boxes = getLeafNode(tree, 2)
elif method_list[k] == 'kd_standard':
tree = Kd_standard(data, param)
tree.buildIndex()
leaf_boxes = getLeafNode(tree, 2)
else:
logging.error('No such index structure!')
sys.exit(1)
workload_counts = [leaf[1] for leaf in leaf_boxes]
res_cube_value[i, j, k] = np.var(workload_counts)
res_value_summary = np.average(res_cube_value, axis=1)
np.savetxt(param.resdir + exp_name + dataset_identifier , res_value_summary, fmt='%.4f\t')
def eval_runtime(data, param):
logging.info("eval_runtime")
exp_name = "eval_runtime"
analyst = 6
param.part_size = analyst
param.ANALYST_COUNT = analyst * analyst
bandwidth = 20 # fixed
method_list = ['grid_standard', 'quad_standard', 'kd_standard']
res_cube_value = np.zeros((2, len(seed_list), len(method_list)))
for j in range(len(seed_list)):
param.seed = seed_list[j]
for k in range(len(method_list)):
start = time.time()
if method_list[k] == 'grid_standard':
tree = Grid_standard(data, param)
elif method_list[k] == 'quad_standard':
tree = Quad_standard(data, param)
elif method_list[k] == 'kd_standard':
tree = Kd_standard(data, param)
else:
logging.error('No such index structure!')
sys.exit(1)
tree.buildIndex()
partition_duration = time.time() - start
answer = optimization(tree, bandwidth, seed_list[j], param)
optimization_duration = time.time() - start - partition_duration
res_cube_value[0, j, k] = partition_duration
res_cube_value[1, j, k] = optimization_duration
res_value_summary = np.average(res_cube_value, axis=1)
np.savetxt(param.resdir + exp_name + dataset_identifier, res_value_summary, fmt='%.4f\t')