class ampligraph.latent_features.ComplEx(k=100, eta=2, epochs=100, batches_count=100, seed=0, embedding_model_params={'corrupt_sides': ['s+o'], 'negative_corruption_entities': 'all'}, optimizer='adam', optimizer_params={'lr': 0.0005}, loss='nll', loss_params={}, regularizer=None, regularizer_params={}, initializer='xavier', initializer_params={'uniform': False}, verbose=False)

Complex embeddings (ComplEx)

The ComplEx model [TWR+16] is an extension of the ampligraph.latent_features.DistMult bilinear diagonal model . ComplEx scoring function is based on the trilinear Hermitian dot product in \(\mathcal{C}\):

\[f_{ComplEx}=Re(\langle \mathbf{r}_p, \mathbf{e}_s, \overline{\mathbf{e}_o} \rangle)\]

ComplEx can be improved if used alongside the nuclear 3-norm (the ComplEx-N3 model [LUO18]) , which can be easily added to the loss function via the regularizer hyperparameter with p=3 and a chosen regularisation weight (represented by lambda), as shown in the example below. See also ampligraph.latent_features.LPRegularizer().


Since ComplEx embeddings belong to \(\mathcal{C}\), this model uses twice as many parameters as ampligraph.latent_features.DistMult.


>>> import numpy as np
>>> from ampligraph.latent_features import ComplEx
>>> model = ComplEx(batches_count=2, seed=555, epochs=100, k=20, eta=5,
>>>             loss='pairwise', loss_params={'margin':1},
>>>             regularizer='LP', regularizer_params={'p': 2, 'lambda':0.1})
>>> X = np.array([['a', 'y', 'b'],
>>>               ['b', 'y', 'a'],
>>>               ['a', 'y', 'c'],
>>>               ['c', 'y', 'a'],
>>>               ['a', 'y', 'd'],
>>>               ['c', 'y', 'd'],
>>>               ['b', 'y', 'c'],
>>>               ['f', 'y', 'e']])
>>> model.predict(np.array([['f', 'y', 'e'], ['b', 'y', 'd']]))
[[0.019520484], [-0.14998421]]
>>> model.get_embeddings(['f','e'], embedding_type='entity')
array([[-0.33021057,  0.26524785,  0.0446662 , -0.07932718, -0.15453218,
    -0.22342539, -0.03382565,  0.17444217,  0.03009969, -0.33569157,
     0.3200497 ,  0.03803705,  0.05536304, -0.00929996,  0.24446663,
     0.34408194,  0.16192885, -0.15033236, -0.19703785, -0.00783876,
     0.1495124 , -0.3578853 , -0.04975723, -0.03930473,  0.1663541 ,
    -0.24731971, -0.141296  ,  0.03150219,  0.15328223, -0.18549544,
    -0.39240393, -0.10824018,  0.03394471, -0.11075485,  0.1367736 ,
     0.10059565, -0.32808647, -0.00472086,  0.14231135, -0.13876757],
   [-0.09483694,  0.3531292 ,  0.04992269, -0.07774793,  0.1635035 ,
     0.30610007,  0.3666711 , -0.13785957, -0.3143734 , -0.36909637,
    -0.13792469, -0.07069954, -0.0368113 , -0.16743314,  0.4090072 ,
    -0.03407392,  0.3113114 , -0.08418448,  0.21435146,  0.12006859,
     0.08447982, -0.02025972,  0.38752195,  0.11451488, -0.0258422 ,
    -0.10990044, -0.22661531, -0.00478273, -0.0238297 , -0.14207476,
     0.11064807,  0.20135397,  0.22501846, -0.1731076 , -0.2770435 ,
     0.30784574, -0.15043163, -0.11599299,  0.05718031, -0.1300622 ]],


__init__([k, eta, epochs, batches_count, …]) Initialize an EmbeddingModel
fit(X[, early_stopping, early_stopping_params]) Train a ComplEx model.
get_embeddings(entities[, embedding_type]) Get the embeddings of entities or relations.
predict(X[, from_idx]) Predict the scores of triples using a trained embedding model.
__init__(k=100, eta=2, epochs=100, batches_count=100, seed=0, embedding_model_params={'corrupt_sides': ['s+o'], 'negative_corruption_entities': 'all'}, optimizer='adam', optimizer_params={'lr': 0.0005}, loss='nll', loss_params={}, regularizer=None, regularizer_params={}, initializer='xavier', initializer_params={'uniform': False}, verbose=False)

Initialize an EmbeddingModel

Also creates a new Tensorflow session for training.

  • k (int) – Embedding space dimensionality
  • eta (int) – The number of negatives that must be generated at runtime during training for each positive.
  • epochs (int) – The iterations of the training loop.
  • batches_count (int) – The number of batches in which the training set must be split during the training loop.
  • seed (int) – The seed used by the internal random numbers generator.
  • embedding_model_params (dict) –

    ComplEx-specific hyperparams:

    • ’negative_corruption_entities’ - Entities to be used for generation of corruptions while training. It can take the following values : all (default: all entities), batch (entities present in each batch), list of entities or an int (which indicates how many entities that should be used for corruption generation).
    • corrupt_sides : Specifies how to generate corruptions for training. Takes values s, o, s+o or any combination passed as a list
  • optimizer (string) – The optimizer used to minimize the loss function. Choose between ‘sgd’, ‘adagrad’, ‘adam’, ‘momentum’.
  • optimizer_params (dict) –

    Arguments specific to the optimizer, passed as a dictionary.

    Supported keys:

    • ’lr’ (float): learning rate (used by all the optimizers). Default: 0.1.
    • ’momentum’ (float): learning momentum (only used when optimizer=momentum). Default: 0.9.

    Example: optimizer_params={'lr': 0.01}

  • loss (string) –

    The type of loss function to use during training.

    • pairwise the model will use pairwise margin-based loss function.
    • nll the model will use negative loss likelihood.
    • absolute_margin the model will use absolute margin likelihood.
    • self_adversarial the model will use adversarial sampling loss function.
    • multiclass_nll the model will use multiclass nll loss. Switch to multiclass loss defined in [aC15] by passing ‘corrupt_sides’ as [‘s’,’o’] to embedding_model_params. To use loss defined in [KBK17] pass ‘corrupt_sides’ as ‘o’ to embedding_model_params.
  • loss_params (dict) –

    Dictionary of loss-specific hyperparameters. See loss functions documentation for additional details.

    Example: optimizer_params={'lr': 0.01} if loss='pairwise'.

  • regularizer (string) –

    The regularization strategy to use with the loss function.

    • None: the model will not use any regularizer (default)
    • ’LP’: the model will use L1, L2 or L3 based on the value of regularizer_params['p'] (see below).
  • regularizer_params (dict) –

    Dictionary of regularizer-specific hyperparameters. See the regularizers documentation for additional details.

    Example: regularizer_params={'lambda': 1e-5, 'p': 2} if regularizer='LP'.

  • initializer (string) –

    The type of initializer to use.

    • normal: The embeddings will be initialized from a normal distribution
    • uniform: The embeddings will be initialized from a uniform distribution
    • xavier: The embeddings will be initialized using xavier strategy (default)
  • initializer_params (dict) –

    Dictionary of initializer-specific hyperparameters. See the initializer documentation for additional details.

    Example: initializer_params={'mean': 0, 'std': 0.001} if initializer='normal'.

  • verbose (bool) – Verbose mode.
fit(X, early_stopping=False, early_stopping_params={})

Train a ComplEx model.

The model is trained on a training set X using the training protocol described in [TWR+16].

  • X (ndarray, shape [n, 3]) – The training triples
  • early_stopping (bool) –

    Flag to enable early stopping (default:False).

    If set to True, the training loop adopts the following early stopping heuristic:

    • The model will be trained regardless of early stopping for burn_in epochs.
    • Every check_interval epochs the method will compute the metric specified in criteria.

    If such metric decreases for stop_interval checks, we stop training early.

    Note the metric is computed on x_valid. This is usually a validation set that you held out.

    Also, because criteria is a ranking metric, it requires generating negatives. Entities used to generate corruptions can be specified, as long as the side(s) of a triple to corrupt. The method supports filtered metrics, by passing an array of positives to x_filter. This will be used to filter the negatives generated on the fly (i.e. the corruptions).


    Keep in mind the early stopping criteria may introduce a certain overhead (caused by the metric computation). The goal is to strike a good trade-off between such overhead and saving training epochs.

    A common approach is to use MRR unfiltered:

    early_stopping_params={x_valid=X['valid'], 'criteria':

    Note the size of validation set also contributes to such overhead. In most cases a smaller validation set would be enough.

  • early_stopping_params (dictionary) –

    Dictionary of hyperparameters for the early stopping heuristics.

    The following string keys are supported:

    • ’x_valid’: ndarray, shape [n, 3] : Validation set to be used for early stopping.
    • ’criteria’: string : criteria for early stopping ‘hits10’, ‘hits3’, ‘hits1’ or ‘mrr’(default).
    • ’x_filter’: ndarray, shape [n, 3] : Positive triples to use as filter if a ‘filtered’ early stopping criteria is desired (i.e. filtered-MRR if ‘criteria’:’mrr’). Note this will affect training time (no filter by default).
    • ’burn_in’: int : Number of epochs to pass before kicking in early stopping (default: 100).
    • check_interval’: int : Early stopping interval after burn-in (default:10).
    • ’stop_interval’: int : Stop if criteria is performing worse over n consecutive checks (default: 3)
    • ’corruption_entities’: List of entities to be used for corruptions. If ‘all’, it uses all entities (default: ‘all’)
    • ’corrupt_side’: Specifies which side to corrupt. ‘s’, ‘o’, ‘s+o’ (default)

    Example: early_stopping_params={x_valid=X['valid'], 'criteria': 'mrr'}

get_embeddings(entities, embedding_type='entity')

Get the embeddings of entities or relations.


Use ampligraph.utils.create_tensorboard_visualizations() to visualize the embeddings with TensorBoard.

  • entities (array-like, dtype=int, shape=[n]) – The entities (or relations) of interest. Element of the vector must be the original string literals, and not internal IDs.
  • embedding_type (string) – If ‘entity’, entities argument will be considered as a list of knowledge graph entities (i.e. nodes). If set to ‘relation’, they will be treated as relation types instead (i.e. predicates).

embeddings – An array of k-dimensional embeddings.

Return type:

ndarray, shape [n, k]

predict(X, from_idx=False)

Predict the scores of triples using a trained embedding model.

The function returns raw scores generated by the model.


To obtain probability estimates, use a logistic sigmoid:

>>> y_pred = model.predict(np.array([['f', 'y', 'e'], ['b', 'y', 'd']]))
>>> print(y_pred)
[-0.31336197, 0.07829369]
>>> from scipy.special import expit
>>> expit(y_pred)
array([0.42229432, 0.51956344], dtype=float32)
  • X (ndarray, shape [n, 3]) – The triples to score.
  • from_idx (bool) – If True, will skip conversion to internal IDs. (default: False).

scores_predict – The predicted scores for input triples X.

Return type:

ndarray, shape [n]