murat karaGloVeThere are works on the statistics of term-term, or term-document relationships in documents in order to generate low-dimensional word…Jun 9Jun 9
murat karaSkip ThoughtsSkip-thoughts is based on the idea that predicting the sentences instead of words as in Skip-Gram model [1]. It is trained with an…Jun 9Jun 9
murat karaTransformers NetworksThis architecture is proposed in “Attention: All You Need” paper [1]. It completely changes the encoder-decoder, i.e. seq-to-seq, model in…Jun 9Jun 9
murat karaSequence-to-Sequence NetworksBefore understanding attention, we need to look at sequence-to-sequence networks. They are based on 2 pioneering papers [1] [2]. They are…Jun 9Jun 9
murat karaRecurrent Neural Networks and LSTMRecurrent Neural Networks allow us to overcome the difficulty of the possibility of processing input of any length. There are two output…Jun 9Jun 9
murat karaConvolutional Neural Networks (CNN)As a specific version of deep learning, CNNs are fine-tuned combinations of convolutional and pooling layers [1].Jun 9Jun 9