The paper "Complex Embeddings for Simple Link Prediction" by Theo Trouillon et al. proposes a novel approach to link prediction in large knowledge bases using complex-valued embeddings. The authors argue that complex embeddings, which use the Hermitian dot product, can effectively handle a wide range of binary relations, including symmetric and antisymmetric relations, while maintaining linear complexity in both space and time. This approach is simpler and more scalable compared to state-of-the-art models like Neural Tensor Networks (NTN) and Holographic Embeddings (HoIE). The paper demonstrates the effectiveness of the proposed method through experiments on synthetic and real datasets, showing superior performance on benchmark link prediction tasks. The authors also provide an equivalent reformulation of their model using only real embeddings, making it easier for practitioners to implement. The paper concludes by discussing potential extensions and future directions for the approach.The paper "Complex Embeddings for Simple Link Prediction" by Theo Trouillon et al. proposes a novel approach to link prediction in large knowledge bases using complex-valued embeddings. The authors argue that complex embeddings, which use the Hermitian dot product, can effectively handle a wide range of binary relations, including symmetric and antisymmetric relations, while maintaining linear complexity in both space and time. This approach is simpler and more scalable compared to state-of-the-art models like Neural Tensor Networks (NTN) and Holographic Embeddings (HoIE). The paper demonstrates the effectiveness of the proposed method through experiments on synthetic and real datasets, showing superior performance on benchmark link prediction tasks. The authors also provide an equivalent reformulation of their model using only real embeddings, making it easier for practitioners to implement. The paper concludes by discussing potential extensions and future directions for the approach.