def test_grid_optimize_6():
"""
Case 6: No xset is given to test. Should raise TypeError.
"""
with pytest.raises(TypeError):
loss = grid_optimizer(X_in, params, args=['n_reservoir'])
return
def test_grid_optimize_7():
"""
Case 7: yset is given, but not xset.
"""
with pytest.raises(TypeError):
loss = grid_optimizer(X_in,
params,
args=['n_reservoir'],
yset=reservoir_set)
return
def test_grid_optimize_3():
"""
Case 3: Two variables specified, only one set given.
"""
with pytest.raises(AssertionError):
loss = grid_optimizer(X_in.T,
params,
args=['n_reservoir', 'sparsity'],
xset=reservoir_set,
verbose=False,
visualize=False)
return
def test_grid_optimize_2():
"""
Case 2: Variable not in params
"""
with pytest.raises(AssertionError):
loss = grid_optimizer(X_in.T,
params,
args=['jimmy'],
xset=trainingLengths,
verbose=False,
visualize=False)
return
def test_grid_optimize_1():
"""
Case 1: Too many variables given
"""
with pytest.raises(AssertionError):
loss = grid_optimizer(X_in.T,
params,
args=['trainlen', 'rho', 'noise'],
xset=trainingLengths,
verbose=False,
visualize=False)
return
def test_grid_optimize_5():
"""
Case 5: ESN does not train because of mismatched input shapes.
"""
with pytest.raises(IndexError):
loss = grid_optimizer(X_in,
params,
args=['n_reservoir'],
xset=reservoir_set,
yset=None,
verbose=False,
visualize=False)
return
def test_grid_optimize_4():
"""
Case 4: One variable specified, two sets given.
"""
with pytest.raises(AssertionError):
loss = grid_optimizer(X_in.T,
params,
args=['n_reservoir'],
xset=reservoir_set,
yset=sparsity_set,
verbose=False,
visualize=False)
return
X_in = np.array(X_in)
print(X_in.shape, len(X_in.shape))
# =============================================================================
# ESN Optimization
# =============================================================================
MAX_TRAINLEN = int(len(xdf) - params['future'])
print(f"Maximum training length is {MAX_TRAINLEN}")
print('Optimizing spectral radius and regularization')
tic = time.perf_counter()
radiusxnoise_loss = grid_optimizer(X_in.T,
params,
args=['rho', 'noise'],
xset=radius_set,
yset=noise_set,
verbose=True,
save_path=save_prefix)
toc = time.perf_counter()
elapsed = toc - tic
print(f"This simulation took {elapsed:0.02f} seconds")
print(f"This simulation took {elapsed/60:0.02f} minutes")
opt_radius, opt_noise = optimal_values(radiusxnoise_loss, radius_set,
noise_set)
params['rho'] = opt_radius
params['noise'] = opt_noise
print('Optimizing network size and sparsity')