Autonomous systems that understand natural language must reason about complex language and visual observations. Key to making progress towards such systems is the availability of benchmark datasets and tasks. We introduce the Cornell Natural Language Visual Reasoning (NLVR) corpus, which targets reasoning skills like counting, comparisons, and set theory. NLVR contains 92,244 examples of natural language statements paired with synthetic images and annotated with boolean values for the simple task of determining whether the sentence is true or false about the image. While it presents a simple task, NLVR has been developed to challenge systems with diverse linguistic phenomena and complex reasoning. Linguistic analysis confirms that NLVR presents diversity and complexity beyond what is provided by contemporary benchmarks. Empirical evaluation of several methods further demonstrates the open challenges NLVR presents.

Linguistic sequence labeling is a general approach encompassing a variety of problems, such as part-of-speech tagging and named entity recognition. Recent advances in neural networks (NNs) make it possible to build reliable models without handcrafted features. However, in many cases, it is hard to obtain sufficient annotations to train these models. In this study, we develop a neural framework to extract knowledge from raw texts and empower the sequence labeling task. Besides word-level knowledge contained in pre-trained word embeddings, character-aware neural language models are incorporated to extract character-level knowledge. Transfer learning techniques are further adopted to mediate different components and guide the language model towards the key knowledge. Comparing to previous methods, these task-specific knowledge allows us to adopt a more concise model and conduct more efficient training. Different from most transfer learning methods, the proposed framework does not rely on any additional supervision. It extracts knowledge from self-contained order information of training sequences. Extensive experiments on benchmark datasets demonstrate the effectiveness of leveraging character-level knowledge and the efficiency of co-training. For example, on the CoNLL03 NER task, model training completes in about 6 hours on a single GPU, reaching F_1 score of 91.71+/-0.10 without using any extra annotations.

Feature representations from pre-trained deep neural networks have been known to exhibit excellent generalization and utility across a variety of related tasks. Fine-tuning is by far the simplest and most widely used approach that seeks to exploit and adapt these feature representations to novel tasks with limited data. Despite the effectiveness of fine-tuning, it is often sub-optimal and requires very careful optimization to prevent severe over-fitting to small datasets. The problem of sub-optimality and overfitting, is due in part to the large number of parameters used in a typical deep convolutional neural network. To address these problems, we propose a simple yet effective regularization method for fine-tuning pre-trained deep networks for the task of k-shot learning. To prevent overfitting, our key strategy is to cluster the model parameters while ensuring intra-cluster similarity and inter-cluster diversity of the parameters, effectively regularizing the dimensionality of the parameter search space. In particular, we identify groups of neurons within each layer of a deep network that shares similar activation patterns. When the network is to be fine-tuned for a classification task using only k examples, we propagate a single gradient to all of the neuron parameters that belong to the same group. The grouping of neurons is non-trivial as neuron activations depend on the distribution of the input data. To efficiently search for optimal groupings conditioned on the input data, we propose a reinforcement learning search strategy using recurrent networks to learn the optimal group assignments for each network layer. Experimental results show that our method can be easily applied to several popular convolutional neural networks and improve upon other state-of-the-art fine-tuning based k-shot learning strategies by more than 10%.

A New AI Evaluation Cosmos: Ready to Play the Game?

We report on a series of new platforms and events dealing with AI evaluation that may change the way in which AI systems are compared and their progress is measured. The introduction of a more diverse and challenging set of tasks in these platforms can feed AI research in the years to come, shaping the notion of success and the directions of the field. However, the playground of tasks and challenges presented there may misdirect the field without some meaningful structure and systematic guidelines for its organization and use. Anticipating this issue, we also report on several initiatives and workshops that are putting the focus on analyzing the similarity and dependencies between tasks, their difficulty, what capabilities they really measure and – ultimately – on elaborating new concepts and tools that can arrange tasks and benchmarks into a meaningful taxonomy.

We release a dataset of 65646 StarCraft replays that contains 1535 million frames and 496 million player actions. We provide full game state data along with the original replays that can be viewed in StarCraft. The game state data was recorded every 3 frames which ensures suitability for a wide variety of machine learning tasks such as strategy classification, inverse reinforcement learning, imitation learning, forward modeling, partial information extraction, and others. We use TorchCraft to extract and store the data, which standardizes the data format for both reading from replays and reading directly from the game. Furthermore, the data can be used on different operating systems and platforms. The dataset contains valid, non-corrupted replays only and its quality and diversity was ensured by a number of heuristics. We illustrate the diversity of the data with various statistics and provide examples of tasks that benefit from the dataset.

Online social media is a social vehicle in which people share various moments of their lives with their friends, such as playing sports, cooking dinner or just taking a selfie for fun, via visual means, that is, photographs. Our study takes a closer look at the popular visual concepts illustrating various cultural lifestyles from aggregated, de-identified photographs. We perform analysis both at macroscopic and microscopic levels, to gain novel insights about global and local visual trends as well as the dynamics of interpersonal cultural exchange and diffusion among Facebook friends. We processed images by automatically classifying the visual content by a convolutional neural network (CNN). Through various statistical tests, we find that socially tied individuals more likely post images showing similar cultural lifestyles. To further identify the main cause of the observed social correlation, we use the Shuffle test and the Preference-based Matched Estimation (PME) test to distinguish the effects of influence and homophily. The results indicate that the visual content of each user's photographs are temporally, although not necessarily causally, correlated with the photographs of their friends, which may suggest the effect of influence. Our paper demonstrates that Facebook photographs exhibit diverse cultural lifestyles and preferences and that the social interaction mediated through the visual channel in social media can be an effective mechanism for cultural diffusion.

Knowledge base(KB) plays an important role in artificial intelligence. Much effort has been taken to both manually and automatically construct web-scale knowledge bases. Comparing with manually constructed KBs, automatically constructed KB is broader but with more noises. In this paper, we study the problem of improving the quality for automatically constructed web-scale knowledge bases, in particular, lexical taxonomies of isA relationships. We find that these taxonomies usually contain cycles, which are often introduced by incorrect isA relations. Inspired by this observation, we introduce two kinds of models to detect incorrect isA relations from cycles. The first one eliminates cycles by extracting directed acyclic graphs, and the other one eliminates cycles by grouping nodes into different levels. We implement our models on Probase, a state-of-the-art, automatically constructed, web-scale taxonomy. After processing tens of millions of relations, our models eliminate 74 thousand wrong relations with 91% accuracy.

Taxonomy is indispensable in understanding natural language. A variety of large scale, usage-based, data-driven lexical taxonomies have been constructed in recent years.Hypernym-hyponym relationship, which is considered as the backbone of lexical taxonomies can not only be used to categorize the data but also enables generalization. In particular, we focus on one of the most prominent properties of the hypernym-hyponym relationship, namely, transitivity, which has a significant implication for many applications. We show that, unlike human crafted ontologies and taxonomies, transitivity does not always hold in data-drivenlexical taxonomies. We introduce a supervised approach to detect whether transitivity holds for any given pair of hypernym-hyponym relationships. Besides solving the inferencing problem, we also use the transitivity to derive new hypernym-hyponym relationships for data-driven lexical taxonomies. We conduct extensive experiments to show the effectiveness of our approach.

Deep Residual Networks (ResNets) have recently achieved state-of-the-art results on many challenging computer vision tasks. In this work we analyze the role of Batch Normalization (BatchNorm) layers on ResNets in the hope of improving the current architecture and better incorporating other normalization techniques, such as Normalization Propagation (NormProp), into ResNets. Firstly, we verify that BatchNorm helps distribute representation learning to residual blocks at all layers, as opposed to a plain ResNet without BatchNorm where learning happens mostly in the latter part of the network. We also observe that BatchNorm well regularizes Concatenated ReLU (CReLU) activation scheme on ResNets, whose magnitude of activation grows by preserving both positive and negative responses when going deeper into the network. Secondly, we investigate the use of NormProp as a replacement for BatchNorm in ResNets. Though NormProp theoretically attains the same effect as BatchNorm on generic convolutional neural networks, the identity mapping of ResNets invalidates its theoretical promise and NormProp exhibits a significant performance drop when naively applied. To bridge the gap between BatchNorm and NormProp in ResNets, we propose a simple modification to NormProp and employ the CReLU activation scheme. We experiment on visual object recognition benchmark datasets such as CIFAR-10/100 and ImageNet and demonstrate that 1) the modified NormProp performs better than the original NormProp but is still not comparable to BatchNorm and 2) CReLU improves the performance of ResNets with or without normalizations.