diff options
Diffstat (limited to 'ml_exp/kernels.py')
| -rw-r--r-- | ml_exp/kernels.py | 85 |
1 files changed, 25 insertions, 60 deletions
diff --git a/ml_exp/kernels.py b/ml_exp/kernels.py index 488232392..11679460b 100644 --- a/ml_exp/kernels.py +++ b/ml_exp/kernels.py @@ -30,15 +30,17 @@ except ImportError: TF_AV = False -def gaussian_kernel(X1, +def laplauss_kernel(X1, X2, sigma, + laplauss='gauss', use_tf=True): """ - Calculates the Gaussian Kernel. + Calculates the Lpalacian or Gaussian Kernel. X1: first representations. X2: second representations. sigma: kernel width. + laplauss: which kernel to calculate. use_tf: if tensorflow should be used. """ # If tf is to be used but couldn't be imported, don't try to use it. @@ -47,62 +49,12 @@ def gaussian_kernel(X1, X1_size = X1.shape[0] X2_size = X2.shape[0] - i_sigma = -0.5 / (sigma*sigma) - - if use_tf: - if tf.config.experimental.list_physical_devices('GPU'): - with tf.device('GPU:0'): - X1 = tf.convert_to_tensor(X1) - X2 = tf.convert_to_tensor(X2) - X2r = tf.rank(X2) - - def cond(i, _): - return tf.less(i, X1_size) - - def body(i, K): - if X2r == 3: - norm = tf.norm(X2 - X1[i], axis=(1, 2)) - else: - norm = tf.norm(X2 - X1[i], axis=-1) - - return (i + 1, - K.write(i, tf.exp(i_sigma * tf.square(norm)))) - - K = tf.TensorArray(dtype=tf.float64, - size=X1_size) - i_state = (0, K) - n, K = tf.while_loop(cond, body, i_state) - K = K.stack() + if laplauss == 'gauss': + i_sigma = -0.5 / (sigma*sigma) + elif laplauss == 'laplace': + i_sigma = -0.5 / sigma else: - K = np.zeros((X1_size, X2_size), dtype=np.float64) - # Faster way of calculating the kernel (no numba support). - for i in range(X1_size): - if X2.ndim == 3: - norm = np.linalg.norm(X2 - X1[i], axis=(1, 2)) - else: - norm = np.linalg.norm(X2 - X1[i], axis=-1) - K[i, :] = np.exp(i_sigma * np.square(norm)) - return K - - -def laplacian_kernel(X1, - X2, - sigma, - use_tf=True): - """ - Calculates the Laplacian Kernel. - X1: first representations. - X2: second representations. - sigma: kernel width. - use_tf: if tensorflow should be used. - """ - # If tf is to be used but couldn't be imported, don't try to use it. - if use_tf and not TF_AV: - use_tf = False - - X1_size = X1.shape[0] - X2_size = X2.shape[0] - i_sigma = -0.5 / sigma + i_sigma = -0.5 / (sigma**2) if use_tf: if tf.config.experimental.list_physical_devices('GPU'): @@ -120,8 +72,15 @@ def laplacian_kernel(X1, else: norm = tf.norm(X2 - X1[i], axis=-1) - return (i + 1, - K.write(i, tf.exp(i_sigma * norm))) + if laplauss == 'gauss': + return (i + 1, + K.write(i, tf.exp(i_sigma * tf.square(norm)))) + elif laplauss == 'laplace': + return (i + 1, + K.write(i, tf.exp(i_sigma * norm))) + else: + return (i + 1, + K.write(i, tf.exp(i_sigma * tf.square(norm)))) K = tf.TensorArray(dtype=tf.float64, size=X1_size) @@ -136,5 +95,11 @@ def laplacian_kernel(X1, norm = np.linalg.norm(X2 - X1[i], axis=(1, 2)) else: norm = np.linalg.norm(X2 - X1[i], axis=-1) - K[i, :] = np.exp(i_sigma * norm) + if laplauuss == 'gauss': + K[i, :] = np.exp(i_sigma * np.square(norm)) + elif laplauce == 'laplace': + K[i, :] = np.exp(i_sigma * norm) + else: + K[i, :] = np.exp(i_sigma * np.square(norm)) + return K |