We present a new neural network architecture which 1) learns word embeddings that better capture the semantics of words by incorporating both local and global document context, and 2) accounts for homonymy and polysemy by learning multiple embeddings per word. Eric H. Huang, Richard Socher, Christopher D. Manning and Andrew Y. Ng in ACL 2012.
We introduce a model based on a combination of convolutional and recursive neural networks (CNN and RNN) for learning features and classifying RGB-D images. Our main result is that even RNNs with random weights compose powerful features. Our model obtains state of the art performance on a standard RGB-D object dataset while being more accurate and faster during training and testing than comparable architectures such as two-layer CNNs.
we develop an end-to-end system for detecting text from natural images. Using a K-means based unsupervised feature learning algorithm and multilayer neural network, we achieve state-of-the-art performance on the Streetview Text and ICDAR 2003 benchmarks.
We develop a recursive neural network (RNN) that learns compositionsl vector representations for phrases and sentences of arbitrary syntactic type and length. In each parse tree node, a vector captures the meaning and a matrix captures how it changes the meaning of neighboring words or phrases.
We consider the problem of building highlevel, class-specific feature detectors from only unlabeled data. For example, is it possible to learn a face detector using only unlabeled images? To answer this, we train a 9layered locally connected sparse autoencoder with pooling and local contrast normalization on a large dataset of images (the model has 1 […]
