We propose a technique for declaratively specifying strategies for semi-supervised learning (SSL). The proposed method can be used to specify ensembles of semi-supervised learning, as well as agreement constraints and entropic regularization constraints between these learners, and can be used to model both well-known heuristics such as co-training and novel domain-specific heuristics. In addition to representing individual SSL heuristics, we show that multiple heuristics can also be automatically combined using Bayesian optimization methods. We show consistent improvements on a suite of well-studied SSL benchmarks, including a new state-of-the-art result on a difficult relation extraction task.

Though impressive results have been achieved in visual captioning, the task of generating abstract stories from photo streams is still a little-tapped problem. Different from captions, stories have more expressive language styles and contain many imaginary concepts that do not appear in the images. Thus it poses challenges to behavioral cloning algorithms. Furthermore, due to the limitations of automatic metrics on evaluating story quality, reinforcement learning methods with hand-crafted rewards also face difficulties in gaining an overall performance boost. Therefore, we propose an Adversarial REward Learning (AREL) framework to learn an implicit reward function from human demonstrations, and then optimize policy search with the learned reward function. Though automatic evaluation indicates slight performance boost over state-of-the-art (SOTA) methods in cloning expert behaviors, human evaluation shows that our approach achieves significant improvement in generating more human-like stories than SOTA systems.

Click through rate (CTR) prediction is very important for Native advertisement but also hard as there is no direct query intent. In this paper we propose a large-scale event embedding scheme to encode the each user browsing event by training a Siamese network with weak supervision on the users' consecutive events. The CTR prediction problem is modeled as a supervised recurrent neural network, which naturally model the user history as a sequence of events. Our proposed recurrent models utilizing pretrained event embedding vectors and an attention layer to model the user history. Our experiments demonstrate that our model significantly outperforms the baseline and some variants.

We study the problem of named entity recognition (NER) from electronic medical records, which is one of the most fundamental and critical problems for medical text mining. Medical records which are written by clinicians from different specialties usually contain quite different terminologies and writing styles. The difference of specialties and the cost of human annotation makes it particularly difficult to train a universal medical NER system. In this paper, we propose a label-aware double transfer learning framework (La-DTL) for cross-specialty NER, so that a medical NER system designed for one specialty could be conveniently applied to another one with minimal annotation efforts. The transferability is guaranteed by 2 components: (i) we propose label-aware MMD for feature representation transfer, and (ii) we perform parameter transfer with a theoretical upper bound which is also label aware. We annotate a new medical NER corpus and conduct extensive experiments on 12 cross-specialty NER tasks. The experimental results demonstrate that La-DTL provides consistent accuracy improvement over strong baselines. Besides, the promising experimental results on non-medical NER scenarios indicate that La-DTL is potential to be seamlessly adapted to a wide range of NER tasks.

We build a virtual agent for learning language in a 2D maze-like world. The agent sees images of the surrounding environment, listens to a virtual teacher, and takes actions to receive rewards. It interactively learns the teacher's language from scratch based on two language use cases: sentence-directed navigation and question answering. It learns simultaneously the visual representations of the world, the language, and the action control. By disentangling language grounding from other computational routines and sharing a concept detection function between language grounding and prediction, the agent reliably interpolates and extrapolates to interpret sentences that contain new word combinations or new words missing from training sentences. The new words are transferred from the answers of language prediction. Such a language ability is trained and evaluated on a population of over 1.6 million distinct sentences consisting of 119 object words, 8 color words, 9 spatial-relation words, and 50 grammatical words. The proposed model significantly outperforms five comparison methods for interpreting zero-shot sentences. In addition, we demonstrate human-interpretable intermediate outputs of the model in the appendix.

A public tax form from OpenAI reveals the crazy salaries of top AI researchers. There are more competitions pushing for improved image recognition models on mobiles, as well as training systems as fast and cheap as possible. Image recognition on mobiles Google has launched another computer vision challenge to push image recognition in real time for mobile phones. The On-Device Visual Intelligence Challenge (ODVI) is part of a workshop track at the Computer Vision and Pattern Recognition conference (CVPR) happening in June later this year in Salt Lake City. It's challenging to build fast, accurate models on small mobile chips given the latency limit.

This is the first in a five-part series exploring the potential of unified deep learning with CPU, GPU and FGPA technologies. This post explores the machine learning potential of combining different advanced technologies. Deep learning and complex machine learning has quickly become one of the most important computationally intensive applications for a wide variety of fields. The combination of large data sets, high-performance computational capabilities, and evolving and improving algorithms has enabled many successful applications which were previously difficult or impossible to consider. This series explores the challenges of deep learning training and inference, and discusses the benefits of a comprehensive approach for combining CPU, GPU, and FPGA technologies, along with the appropriate software frameworks in a unified deep learning architecture.

To compete in today's data-driven world, organizations need to accelerate the digital transformation process that puts technology at the heart of products, services and operations. Digital transformation enables both private and public entities to provide better outcomes and experiences for the people they serve -- from smarter vehicles to personalized healthcare, from customized shopping experiences to the prevention of credit card fraud. A common thread to these and countless other digital transformation use cases is artificial intelligence. AI applications and their underlying technologies, including machine learning and deep learning, enable organizations to train systems to use massive amounts of data to sense, learn, reason, make predictions and evolve. Under the hood, the engine that makes it all go is the blazingly fast processing power of high-performance computing (HPC) clusters.

Today, we're going to take this trained Keras model and deploy it to an iPhone and iOS app using what Apple has dubbed "CoreML", an easy-to-use machine learning framework for Apple applications: My goal today is to show you how simple it is to deploy your Keras model to your iPhone and iOS using CoreML. To be clear, I'm not a mobile developer by any stretch of the imagination, and if I can do it, I'm confident you can do it as well. Feel free to use the code in today's post as a starting point for your own application. But personally, I'm going to continue the theme of this series and build a Pokedex. A Pokedex is a device that exists in the world of Pokemon, a popular TV show, video game, and trading card series (I was/still am a huge Pokemon nerd). Using a Pokedex you can take a picture of a Pokemon (animal-like creatures that exist in the world of Pokemon) and the Pokedex will automatically identify the creature for for you, providing useful information and statistics, such as the Pokemon's height, weight, and any special abilities it may have. You can see an example of a Pokedex in action at the top of this blog post, but again, feel free to swap out my Keras model for your own -- the process is quite simple and straightforward as you'll see later in this guide.

Google Research engineers have developed a deep learning system that can separate voices from audio-visual data recorded in crowded environments. The system they developed is the equivalent of the "cocktail party" effect, a feature of the human brain that can isolate and focus on one or more particular voices in a crowd. The system is designed to work with both audio and video data at the same time. Google says it created its novel tech by feeding it over 100,000 high-quality videos of lectures and talks hosted on YouTube. All talks were given by a single speaker, with minimal background noise. They trained the AI to recognize sounds based on lip/mouth movement.