Kevin Murphy

Lijun Yu, Yong Cheng, Zhiruo Wang et al. · cmu, deepmind

In this work, we introduce Semantic Pyramid AutoEncoder (SPAE) for enabling frozen LLMs to perform both understanding and generation tasks involving non-linguistic modalities such as images or videos. SPAE converts between raw pixels and interpretable lexical tokens (or words) extracted from the LLM's vocabulary. The resulting tokens capture both the semantic meaning and the fine-grained details needed for visual reconstruction, effectively translating the visual content into a language comprehensible to the LLM, and empowering it to perform a wide array of multimodal tasks. Our approach is validated through in-context learning experiments with frozen PaLM 2 and GPT 3.5 on a diverse set of image understanding and generation tasks. Our method marks the first successful attempt to enable a frozen LLM to generate image content while surpassing state-of-the-art performance in image understanding tasks, under the same setting, by over 25%.

Kevin Murphy

We present BLF (Bayesian Linguistic Forecaster), an agentic system for binary forecasting that achieves state-of-the-art performance on the ForecastBench benchmark. The system is built on three ideas. (1) A linguistic belief state: a semi-structured representation combining numerical probability estimates with natural-language evidence summaries, updated by the LLM at each step of an iterative tool-use loop. This contrasts with the common approach of appending all retrieved evidence to an ever-growing context. (2) Hierarchical multi-trial aggregation: running $K$ independent trials and combining them using logit-space shrinkage with a data-dependent prior. (3) Hierarchical calibration: Platt scaling with a hierarchical prior, which avoids over-shrinking extreme predictions for sources with skewed base rates. On 400 backtesting questions from the ForecastBench leaderboard, BLF outperforms all the top public methods, including Cassi, GPT-5, Grok~4.20, and Foresight-32B. Ablation studies show that the structured belief state is almost as impactful as web search access, and that shrinkage aggregation and hierarchical calibration each provide significant additional gains. In addition, we develop a robust back-testing framework with a leakage rate below 1.5\%, and use rigorous statistical methodology to compare different methods while controlling for various sources of noise.

Yiding Jiang, Shixiang Gu, Kevin Murphy et al.

Solving complex, temporally-extended tasks is a long-standing problem in reinforcement learning (RL). We hypothesize that one critical element of solving such problems is the notion of compositionality. With the ability to learn concepts and sub-skills that can be composed to solve longer tasks, i.e. hierarchical RL, we can acquire temporally-extended behaviors. However, acquiring effective yet general abstractions for hierarchical RL is remarkably challenging. In this paper, we propose to use language as the abstraction, as it provides unique compositional structure, enabling fast learning and combinatorial generalization, while retaining tremendous flexibility, making it suitable for a variety of problems. Our approach learns an instruction-following low-level policy and a high-level policy that can reuse abstractions across tasks, in essence, permitting agents to reason using structured language. To study compositional task learning, we introduce an open-source object interaction environment built using the MuJoCo physics engine and the CLEVR engine. We find that, using our approach, agents can learn to solve to diverse, temporally-extended tasks such as object sorting and multi-object rearrangement, including from raw pixel observations. Our analysis reveals that the compositional nature of language is critical for learning diverse sub-skills and systematically generalizing to new sub-skills in comparison to non-compositional abstractions that use the same supervision.

Junhua Mao, Jonathan Huang, Alexander Toshev et al.

We propose a method that can generate an unambiguous description (known as a referring expression) of a specific object or region in an image, and which can also comprehend or interpret such an expression to infer which object is being described. We show that our method outperforms previous methods that generate descriptions of objects without taking into account other potentially ambiguous objects in the scene. Our model is inspired by recent successes of deep learning methods for image captioning, but while image captioning is difficult to evaluate, our task allows for easy objective evaluation. We also present a new large-scale dataset for referring expressions, based on MS-COCO. We have released the dataset and a toolbox for visualization and evaluation, see https://github.com/mjhucla/Google_Refexp_toolbox

Antoine Dedieu, Joseph Ortiz, Xinghua Lou et al.

We present three improvements to the standard model-based RL paradigm based on transformers: (a) "Dyna with warmup", which trains the policy on real and imaginary data, but only starts using imaginary data after the world model has been sufficiently trained; (b) "nearest neighbor tokenizer" for image patches, which improves upon previous tokenization schemes, which are needed when using a transformer world model (TWM), by ensuring the code words are static after creation, thus providing a constant target for TWM learning; and (c) "block teacher forcing", which allows the TWM to reason jointly about the future tokens of the next timestep, instead of generating them sequentially. We then show that our method significantly improves upon prior methods in various environments. We mostly focus on the challenging Craftax-classic benchmark, where our method achieves a reward of 69.66% after only 1M environment steps, significantly outperforming DreamerV3, which achieves 53.2%, and exceeding human performance of 65.0% for the first time. We also show preliminary results on Craftax-full, MinAtar, and three different two-player games, to illustrate the generality of the approach.

Sireesh Gururaja, Yueheng Zhang, Guannan Tang et al. · cmu

Recent years in NLP have seen the continued development of domain-specific information extraction tools for scientific documents, alongside the release of increasingly multimodal pretrained transformer models. While the opportunity for scientists outside of NLP to evaluate and apply such systems to their own domains has never been clearer, these models are difficult to compare: they accept different input formats, are often black-box and give little insight into processing failures, and rarely handle PDF documents, the most common format of scientific publication. In this work, we present Collage, a tool designed for rapid prototyping, visualization, and evaluation of different information extraction models on scientific PDFs. Collage allows the use and evaluation of any HuggingFace token classifier, several LLMs, and multiple other task-specific models out of the box, and provides extensible software interfaces to accelerate experimentation with new models. Further, we enable both developers and users of NLP-based tools to inspect, debug, and better understand modeling pipelines by providing granular views of intermediate states of processing. We demonstrate our system in the context of information extraction to assist with literature review in materials science.

Xinghua Lou, Meet Dave, Shrinu Kushagra et al.

We propose an approach to simulating trajectories of multiple interacting agents (road users) based on transformers and probabilistic graphical models (PGMs), and apply it to the Waymo SimAgents challenge. The transformer baseline is based on the MTR model, which predicts multiple future trajectories conditioned on the past trajectories and static road layout features. We then improve upon these generated trajectories using a PGM, which contains factors which encode prior knowledge, such as a preference for smooth trajectories, and avoidance of collisions with static obstacles and other moving agents. We perform (approximate) MAP inference in this PGM using the Gauss-Newton method. Finally we sample $K=32$ trajectories for each of the $N \sim 100$ agents for the next $T=8 Δ$ time steps, where $Δ=10$ is the sampling rate per second. Following the Model Predictive Control (MPC) paradigm, we only return the first element of our forecasted trajectories at each step, and then we replan, so that the simulation can constantly adapt to its changing environment. We therefore call our approach "Model Predictive Simulation" or MPS. We show that MPS improves upon the MTR baseline, especially in safety critical metrics such as collision rate. Furthermore, our approach is compatible with any underlying forecasting model, and does not require extra training, so we believe it is a valuable contribution to the community.

Christof Angermueller, David Belanger, Andreea Gane et al.

The use of black-box optimization for the design of new biological sequences is an emerging research area with potentially revolutionary impact. The cost and latency of wet-lab experiments requires methods that find good sequences in few experimental rounds of large batches of sequences--a setting that off-the-shelf black-box optimization methods are ill-equipped to handle. We find that the performance of existing methods varies drastically across optimization tasks, posing a significant obstacle to real-world applications. To improve robustness, we propose Population-Based Black-Box Optimization (P3BO), which generates batches of sequences by sampling from an ensemble of methods. The number of sequences sampled from any method is proportional to the quality of sequences it previously proposed, allowing P3BO to combine the strengths of individual methods while hedging against their innate brittleness. Adapting the hyper-parameters of each of the methods online using evolutionary optimization further improves performance. Through extensive experiments on in-silico optimization tasks, we show that P3BO outperforms any single method in its population, proposing higher quality sequences as well as more diverse batches. As such, P3BO and Adaptive-P3BO are a crucial step towards deploying ML to real-world sequence design.

Ines Chami, Sami Abu-El-Haija, Bryan Perozzi et al.

There has been a surge of recent interest in learning representations for graph-structured data. Graph representation learning methods have generally fallen into three main categories, based on the availability of labeled data. The first, network embedding (such as shallow graph embedding or graph auto-encoders), focuses on learning unsupervised representations of relational structure. The second, graph regularized neural networks, leverages graphs to augment neural network losses with a regularization objective for semi-supervised learning. The third, graph neural networks, aims to learn differentiable functions over discrete topologies with arbitrary structure. However, despite the popularity of these areas there has been surprisingly little work on unifying the three paradigms. Here, we aim to bridge the gap between graph neural networks, network embedding and graph regularization models. We propose a comprehensive taxonomy of representation learning methods for graph-structured data, aiming to unify several disparate bodies of work. Specifically, we propose a Graph Encoder Decoder Model (GRAPHEDM), which generalizes popular algorithms for semi-supervised learning on graphs (e.g. GraphSage, Graph Convolutional Networks, Graph Attention Networks), and unsupervised learning of graph representations (e.g. DeepWalk, node2vec, etc) into a single consistent approach. To illustrate the generality of this approach, we fit over thirty existing methods into this framework. We believe that this unifying view both provides a solid foundation for understanding the intuition behind these methods, and enables future research in the area.

Abhishek Kumar, Ben Poole, Kevin Murphy

Invertible flow-based generative models are an effective method for learning to generate samples, while allowing for tractable likelihood computation and inference. However, the invertibility requirement restricts models to have the same latent dimensionality as the inputs. This imposes significant architectural, memory, and computational costs, making them more challenging to scale than other classes of generative models such as Variational Autoencoders (VAEs). We propose a generative model based on probability flows that does away with the bijectivity requirement on the model and only assumes injectivity. This also provides another perspective on regularized autoencoders (RAEs), with our final objectives resembling RAEs with specific regularizers that are derived by lower bounding the probability flow objective. We empirically demonstrate the promise of the proposed model, improving over VAEs and AEs in terms of sample quality.

Junwei Liang, Lu Jiang, Kevin Murphy et al.

This paper studies the problem of predicting the distribution over multiple possible future paths of people as they move through various visual scenes. We make two main contributions. The first contribution is a new dataset, created in a realistic 3D simulator, which is based on real world trajectory data, and then extrapolated by human annotators to achieve different latent goals. This provides the first benchmark for quantitative evaluation of the models to predict multi-future trajectories. The second contribution is a new model to generate multiple plausible future trajectories, which contains novel designs of using multi-scale location encodings and convolutional RNNs over graphs. We refer to our model as Multiverse. We show that our model achieves the best results on our dataset, as well as on the real-world VIRAT/ActEV dataset (which just contains one possible future).

Zhe Dong, Bryan A. Seybold, Kevin P. Murphy et al.

We propose an efficient inference method for switching nonlinear dynamical systems. The key idea is to learn an inference network which can be used as a proposal distribution for the continuous latent variables, while performing exact marginalization of the discrete latent variables. This allows us to use the reparameterization trick, and apply end-to-end training with stochastic gradient descent. We show that the proposed method can successfully segment time series data, including videos and 3D human pose, into meaningful ``regimes'' by using the piece-wise nonlinear dynamics.

Matthias Minderer, Chen Sun, Ruben Villegas et al.

Extracting and predicting object structure and dynamics from videos without supervision is a major challenge in machine learning. To address this challenge, we adopt a keypoint-based image representation and learn a stochastic dynamics model of the keypoints. Future frames are reconstructed from the keypoints and a reference frame. By modeling dynamics in the keypoint coordinate space, we achieve stable learning and avoid compounding of errors in pixel space. Our method improves upon unstructured representations both for pixel-level video prediction and for downstream tasks requiring object-level understanding of motion dynamics. We evaluate our model on diverse datasets: a multi-agent sports dataset, the Human3.6M dataset, and datasets based on continuous control tasks from the DeepMind Control Suite. The spatially structured representation outperforms unstructured representations on a range of motion-related tasks such as object tracking, action recognition and reward prediction.

Chen Sun, Fabien Baradel, Kevin Murphy et al.

This paper proposes a self-supervised learning approach for video features that results in significantly improved performance on downstream tasks (such as video classification, captioning and segmentation) compared to existing methods. Our method extends the BERT model for text sequences to the case of sequences of real-valued feature vectors, by replacing the softmax loss with noise contrastive estimation (NCE). We also show how to learn representations from sequences of visual features and sequences of words derived from ASR (automatic speech recognition), and show that such cross-modal training (when possible) helps even more.

Chen Sun, Abhinav Shrivastava, Carl Vondrick et al.

This paper focuses on multi-person action forecasting in videos. More precisely, given a history of H previous frames, the goal is to detect actors and to predict their future actions for the next T frames. Our approach jointly models temporal and spatial interactions among different actors by constructing a recurrent graph, using actor proposals obtained with Faster R-CNN as nodes. Our method learns to select a subset of discriminative relations without requiring explicit supervision, thus enabling us to tackle challenging visual data. We refer to our model as Discriminative Relational Recurrent Network (DRRN). Evaluation of action prediction on AVA demonstrates the effectiveness of our proposed method compared to simpler baselines. Furthermore, we significantly improve performance on the task of early action classification on J-HMDB, from the previous SOTA of 48% to 60%.

Chen Sun, Austin Myers, Carl Vondrick et al.

Self-supervised learning has become increasingly important to leverage the abundance of unlabeled data available on platforms like YouTube. Whereas most existing approaches learn low-level representations, we propose a joint visual-linguistic model to learn high-level features without any explicit supervision. In particular, inspired by its recent success in language modeling, we build upon the BERT model to learn bidirectional joint distributions over sequences of visual and linguistic tokens, derived from vector quantization of video data and off-the-shelf speech recognition outputs, respectively. We use VideoBERT in numerous tasks, including action classification and video captioning. We show that it can be applied directly to open-vocabulary classification, and confirm that large amounts of training data and cross-modal information are critical to performance. Furthermore, we outperform the state-of-the-art on video captioning, and quantitative results verify that the model learns high-level semantic features.

Zhenjia Xu, Zhijian Liu, Chen Sun et al.

Humans easily recognize object parts and their hierarchical structure by watching how they move; they can then predict how each part moves in the future. In this paper, we propose a novel formulation that simultaneously learns a hierarchical, disentangled object representation and a dynamics model for object parts from unlabeled videos. Our Parts, Structure, and Dynamics (PSD) model learns to, first, recognize the object parts via a layered image representation; second, predict hierarchy via a structural descriptor that composes low-level concepts into a hierarchical structure; and third, model the system dynamics by predicting the future. Experiments on multiple real and synthetic datasets demonstrate that our PSD model works well on all three tasks: segmenting object parts, building their hierarchical structure, and capturing their motion distributions.

Chris Ying, Aaron Klein, Esteban Real et al.

Recent advances in neural architecture search (NAS) demand tremendous computational resources, which makes it difficult to reproduce experiments and imposes a barrier-to-entry to researchers without access to large-scale computation. We aim to ameliorate these problems by introducing NAS-Bench-101, the first public architecture dataset for NAS research. To build NAS-Bench-101, we carefully constructed a compact, yet expressive, search space, exploiting graph isomorphisms to identify 423k unique convolutional architectures. We trained and evaluated all of these architectures multiple times on CIFAR-10 and compiled the results into a large dataset of over 5 million trained models. This allows researchers to evaluate the quality of a diverse range of models in milliseconds by querying the pre-computed dataset. We demonstrate its utility by analyzing the dataset as a whole and by benchmarking a range of architecture optimization algorithms.

Nam Vo, Lu Jiang, Chen Sun et al.

In this paper, we study the task of image retrieval, where the input query is specified in the form of an image plus some text that describes desired modifications to the input image. For example, we may present an image of the Eiffel tower, and ask the system to find images which are visually similar but are modified in small ways, such as being taken at nighttime instead of during the day. To tackle this task, we learn a similarity metric between a target image and a source image plus source text, an embedding and composing function such that target image feature is close to the source image plus text composition feature. We propose a new way to combine image and text using such function that is designed for the retrieval task. We show this outperforms existing approaches on 3 different datasets, namely Fashion-200k, MIT-States and a new synthetic dataset we create based on CLEVR. We also show that our approach can be used to classify input queries, in addition to image retrieval.

Seong Joon Oh, Kevin Murphy, Jiyan Pan et al.

Instance embeddings are an efficient and versatile image representation that facilitates applications like recognition, verification, retrieval, and clustering. Many metric learning methods represent the input as a single point in the embedding space. Often the distance between points is used as a proxy for match confidence. However, this can fail to represent uncertainty arising when the input is ambiguous, e.g., due to occlusion or blurriness. This work addresses this issue and explicitly models the uncertainty by hedging the location of each input in the embedding space. We introduce the hedged instance embedding (HIB) in which embeddings are modeled as random variables and the model is trained under the variational information bottleneck principle. Empirical results on our new N-digit MNIST dataset show that our method leads to the desired behavior of hedging its bets across the embedding space upon encountering ambiguous inputs. This results in improved performance for image matching and classification tasks, more structure in the learned embedding space, and an ability to compute a per-exemplar uncertainty measure that is correlated with downstream performance.

Chen Sun, Abhinav Shrivastava, Carl Vondrick et al.

Current state-of-the-art approaches for spatio-temporal action localization rely on detections at the frame level and model temporal context with 3D ConvNets. Here, we go one step further and model spatio-temporal relations to capture the interactions between human actors, relevant objects and scene elements essential to differentiate similar human actions. Our approach is weakly supervised and mines the relevant elements automatically with an actor-centric relational network (ACRN). ACRN computes and accumulates pair-wise relation information from actor and global scene features, and generates relation features for action classification. It is implemented as neural networks and can be trained jointly with an existing action detection system. We show that ACRN outperforms alternative approaches which capture relation information, and that the proposed framework improves upon the state-of-the-art performance on JHMDB and AVA. A visualization of the learned relation features confirms that our approach is able to attend to the relevant relations for each action.

Carl Vondrick, Abhinav Shrivastava, Alireza Fathi et al.

We use large amounts of unlabeled video to learn models for visual tracking without manual human supervision. We leverage the natural temporal coherency of color to create a model that learns to colorize gray-scale videos by copying colors from a reference frame. Quantitative and qualitative experiments suggest that this task causes the model to automatically learn to track visual regions. Although the model is trained without any ground-truth labels, our method learns to track well enough to outperform the latest methods based on optical flow. Moreover, our results suggest that failures to track are correlated with failures to colorize, indicating that advancing video colorization may further improve self-supervised visual tracking.

Saining Xie, Chen Sun, Jonathan Huang et al.

Despite the steady progress in video analysis led by the adoption of convolutional neural networks (CNNs), the relative improvement has been less drastic as that in 2D static image classification. Three main challenges exist including spatial (image) feature representation, temporal information representation, and model/computation complexity. It was recently shown by Carreira and Zisserman that 3D CNNs, inflated from 2D networks and pretrained on ImageNet, could be a promising way for spatial and temporal representation learning. However, as for model/computation complexity, 3D CNNs are much more expensive than 2D CNNs and prone to overfit. We seek a balance between speed and accuracy by building an effective and efficient video classification system through systematic exploration of critical network design choices. In particular, we show that it is possible to replace many of the 3D convolutions by low-cost 2D convolutions. Rather surprisingly, best result (in both speed and accuracy) is achieved when replacing the 3D convolutions at the bottom of the network, suggesting that temporal representation learning on high-level semantic features is more useful. Our conclusion generalizes to datasets with very different properties. When combined with several other cost-effective designs including separable spatial/temporal convolution and feature gating, our system results in an effective video classification system that that produces very competitive results on several action classification benchmarks (Kinetics, Something-something, UCF101 and HMDB), as well as two action detection (localization) benchmarks (JHMDB and UCF101-24).

Chenxi Liu, Barret Zoph, Maxim Neumann et al.

We propose a new method for learning the structure of convolutional neural networks (CNNs) that is more efficient than recent state-of-the-art methods based on reinforcement learning and evolutionary algorithms. Our approach uses a sequential model-based optimization (SMBO) strategy, in which we search for structures in order of increasing complexity, while simultaneously learning a surrogate model to guide the search through structure space. Direct comparison under the same search space shows that our method is up to 5 times more efficient than the RL method of Zoph et al. (2018) in terms of number of models evaluated, and 8 times faster in terms of total compute. The structures we discover in this way achieve state of the art classification accuracies on CIFAR-10 and ImageNet.

Amélie Royer, Konstantinos Bousmalis, Stephan Gouws et al.

Style transfer usually refers to the task of applying color and texture information from a specific style image to a given content image while preserving the structure of the latter. Here we tackle the more generic problem of semantic style transfer: given two unpaired collections of images, we aim to learn a mapping between the corpus-level style of each collection, while preserving semantic content shared across the two domains. We introduce XGAN ("Cross-GAN"), a dual adversarial autoencoder, which captures a shared representation of the common domain semantic content in an unsupervised way, while jointly learning the domain-to-domain image translations in both directions. We exploit ideas from the domain adaptation literature and define a semantic consistency loss which encourages the model to preserve semantics in the learned embedding space. We report promising qualitative results for the task of face-to-cartoon translation. The cartoon dataset, CartoonSet, we collected for this purpose is publicly available at google.github.io/cartoonset/ as a new benchmark for semantic style transfer.

Alexander A. Alemi, Ben Poole, Ian Fischer et al.

Recent work in unsupervised representation learning has focused on learning deep directed latent-variable models. Fitting these models by maximizing the marginal likelihood or evidence is typically intractable, thus a common approximation is to maximize the evidence lower bound (ELBO) instead. However, maximum likelihood training (whether exact or approximate) does not necessarily result in a good latent representation, as we demonstrate both theoretically and empirically. In particular, we derive variational lower and upper bounds on the mutual information between the input and the latent variable, and use these bounds to derive a rate-distortion curve that characterizes the tradeoff between compression and reconstruction accuracy. Using this framework, we demonstrate that there is a family of models with identical ELBO, but different quantitative and qualitative characteristics. Our framework also suggests a simple new method to ensure that latent variable models with powerful stochastic decoders do not ignore their latent code.

Ramakrishna Vedantam, Ian Fischer, Jonathan Huang et al.

It is easy for people to imagine what a man with pink hair looks like, even if they have never seen such a person before. We call the ability to create images of novel semantic concepts visually grounded imagination. In this paper, we show how we can modify variational auto-encoders to perform this task. Our method uses a novel training objective, and a novel product-of-experts inference network, which can handle partially specified (abstract) concepts in a principled and efficient way. We also propose a set of easy-to-compute evaluation metrics that capture our intuitive notions of what it means to have good visual imagination, namely correctness, coverage, and compositionality (the 3 C's). Finally, we perform a detailed comparison of our method with two existing joint image-attribute VAE methods (the JMVAE method of Suzuki et.al. and the BiVCCA method of Wang et.al.) by applying them to two datasets: the MNIST-with-attributes dataset (which we introduce here), and the CelebA dataset.

Sergio Guadarrama, Ryan Dahl, David Bieber et al.

We propose a novel approach to automatically produce multiple colorized versions of a grayscale image. Our method results from the observation that the task of automated colorization is relatively easy given a low-resolution version of the color image. We first train a conditional PixelCNN to generate a low resolution color for a given grayscale image. Then, given the generated low-resolution color image and the original grayscale image as inputs, we train a second CNN to generate a high-resolution colorization of an image. We demonstrate that our approach produces more diverse and plausible colorizations than existing methods, as judged by human raters in a "Visual Turing Test".

Alireza Fathi, Zbigniew Wojna, Vivek Rathod et al.

We propose a new method for semantic instance segmentation, by first computing how likely two pixels are to belong to the same object, and then by grouping similar pixels together. Our similarity metric is based on a deep, fully convolutional embedding model. Our grouping method is based on selecting all points that are sufficiently similar to a set of "seed points", chosen from a deep, fully convolutional scoring model. We show competitive results on the Pascal VOC instance segmentation benchmark.

Ramakrishna Vedantam, Samy Bengio, Kevin Murphy et al.

We introduce an inference technique to produce discriminative context-aware image captions (captions that describe differences between images or visual concepts) using only generic context-agnostic training data (captions that describe a concept or an image in isolation). For example, given images and captions of "siamese cat" and "tiger cat", we generate language that describes the "siamese cat" in a way that distinguishes it from "tiger cat". Our key novelty is that we show how to do joint inference over a language model that is context-agnostic and a listener which distinguishes closely-related concepts. We first apply our technique to a justification task, namely to describe why an image contains a particular fine-grained category as opposed to another closely-related category of the CUB-200-2011 dataset. We then study discriminative image captioning to generate language that uniquely refers to one of two semantically-similar images in the COCO dataset. Evaluations with discriminative ground truth for justification and human studies for discriminative image captioning reveal that our approach outperforms baseline generative and speaker-listener approaches for discrimination.

George Papandreou, Tyler Zhu, Nori Kanazawa et al.

We propose a method for multi-person detection and 2-D pose estimation that achieves state-of-art results on the challenging COCO keypoints task. It is a simple, yet powerful, top-down approach consisting of two stages. In the first stage, we predict the location and scale of boxes which are likely to contain people; for this we use the Faster RCNN detector. In the second stage, we estimate the keypoints of the person potentially contained in each proposed bounding box. For each keypoint type we predict dense heatmaps and offsets using a fully convolutional ResNet. To combine these outputs we introduce a novel aggregation procedure to obtain highly localized keypoint predictions. We also use a novel form of keypoint-based Non-Maximum-Suppression (NMS), instead of the cruder box-level NMS, and a novel form of keypoint-based confidence score estimation, instead of box-level scoring. Trained on COCO data alone, our final system achieves average precision of 0.649 on the COCO test-dev set and the 0.643 test-standard sets, outperforming the winner of the 2016 COCO keypoints challenge and other recent state-of-art. Further, by using additional in-house labeled data we obtain an even higher average precision of 0.685 on the test-dev set and 0.673 on the test-standard set, more than 5% absolute improvement compared to the previous best performing method on the same dataset.

Hyun Oh Song, Stefanie Jegelka, Vivek Rathod et al.

Learning the representation and the similarity metric in an end-to-end fashion with deep networks have demonstrated outstanding results for clustering and retrieval. However, these recent approaches still suffer from the performance degradation stemming from the local metric training procedure which is unaware of the global structure of the embedding space. We propose a global metric learning scheme for optimizing the deep metric embedding with the learnable clustering function and the clustering metric (NMI) in a novel structured prediction framework. Our experiments on CUB200-2011, Cars196, and Stanford online products datasets show state of the art performance both on the clustering and retrieval tasks measured in the NMI and Recall@K evaluation metrics.

Alexander A. Alemi, Ian Fischer, Joshua V. Dillon et al.

We present a variational approximation to the information bottleneck of Tishby et al. (1999). This variational approach allows us to parameterize the information bottleneck model using a neural network and leverage the reparameterization trick for efficient training. We call this method "Deep Variational Information Bottleneck", or Deep VIB. We show that models trained with the VIB objective outperform those that are trained with other forms of regularization, in terms of generalization performance and robustness to adversarial attack.

Jonathan Huang, Vivek Rathod, Chen Sun et al.

The goal of this paper is to serve as a guide for selecting a detection architecture that achieves the right speed/memory/accuracy balance for a given application and platform. To this end, we investigate various ways to trade accuracy for speed and memory usage in modern convolutional object detection systems. A number of successful systems have been proposed in recent years, but apples-to-apples comparisons are difficult due to different base feature extractors (e.g., VGG, Residual Networks), different default image resolutions, as well as different hardware and software platforms. We present a unified implementation of the Faster R-CNN [Ren et al., 2015], R-FCN [Dai et al., 2016] and SSD [Liu et al., 2015] systems, which we view as "meta-architectures" and trace out the speed/accuracy trade-off curve created by using alternative feature extractors and varying other critical parameters such as image size within each of these meta-architectures. On one extreme end of this spectrum where speed and memory are critical, we present a detector that achieves real time speeds and can be deployed on a mobile device. On the opposite end in which accuracy is critical, we present a detector that achieves state-of-the-art performance measured on the COCO detection task.

Liang-Chieh Chen, George Papandreou, Iasonas Kokkinos et al.

In this work we address the task of semantic image segmentation with Deep Learning and make three main contributions that are experimentally shown to have substantial practical merit. First, we highlight convolution with upsampled filters, or 'atrous convolution', as a powerful tool in dense prediction tasks. Atrous convolution allows us to explicitly control the resolution at which feature responses are computed within Deep Convolutional Neural Networks. It also allows us to effectively enlarge the field of view of filters to incorporate larger context without increasing the number of parameters or the amount of computation. Second, we propose atrous spatial pyramid pooling (ASPP) to robustly segment objects at multiple scales. ASPP probes an incoming convolutional feature layer with filters at multiple sampling rates and effective fields-of-views, thus capturing objects as well as image context at multiple scales. Third, we improve the localization of object boundaries by combining methods from DCNNs and probabilistic graphical models. The commonly deployed combination of max-pooling and downsampling in DCNNs achieves invariance but has a toll on localization accuracy. We overcome this by combining the responses at the final DCNN layer with a fully connected Conditional Random Field (CRF), which is shown both qualitatively and quantitatively to improve localization performance. Our proposed "DeepLab" system sets the new state-of-art at the PASCAL VOC-2012 semantic image segmentation task, reaching 79.7% mIOU in the test set, and advances the results on three other datasets: PASCAL-Context, PASCAL-Person-Part, and Cityscapes. All of our code is made publicly available online.

Jonathan Huang, Kevin Murphy

We present a generative model of images based on layering, in which image layers are individually generated, then composited from front to back. We are thus able to factor the appearance of an image into the appearance of individual objects within the image --- and additionally for each individual object, we can factor content from pose. Unlike prior work on layered models, we learn a shape prior for each object/layer, allowing the model to tease out which object is in front by looking for a consistent shape, without needing access to motion cues or any labeled data. We show that ordinary stochastic gradient variational bayes (SGVB), which optimizes our fully differentiable lower-bound on the log-likelihood, is sufficient to learn an interpretable representation of images. Finally we present experiments demonstrating the effectiveness of the model for inferring foreground and background objects in images.

Liang-Chieh Chen, Jonathan T. Barron, George Papandreou et al.

Deep convolutional neural networks (CNNs) are the backbone of state-of-art semantic image segmentation systems. Recent work has shown that complementing CNNs with fully-connected conditional random fields (CRFs) can significantly enhance their object localization accuracy, yet dense CRF inference is computationally expensive. We propose replacing the fully-connected CRF with domain transform (DT), a modern edge-preserving filtering method in which the amount of smoothing is controlled by a reference edge map. Domain transform filtering is several times faster than dense CRF inference and we show that it yields comparable semantic segmentation results, accurately capturing object boundaries. Importantly, our formulation allows learning the reference edge map from intermediate CNN features instead of using the image gradient magnitude as in standard DT filtering. This produces task-specific edges in an end-to-end trainable system optimizing the target semantic segmentation quality.

Vignesh Ramanathan, Jonathan Huang, Sami Abu-El-Haija et al.

Multi-person event recognition is a challenging task, often with many people active in the scene but only a small subset contributing to an actual event. In this paper, we propose a model which learns to detect events in such videos while automatically "attending" to the people responsible for the event. Our model does not use explicit annotations regarding who or where those people are during training and testing. In particular, we track people in videos and use a recurrent neural network (RNN) to represent the track features. We learn time-varying attention weights to combine these features at each time-instant. The attended features are then processed using another RNN for event detection/classification. Since most video datasets with multiple people are restricted to a small number of videos, we also collected a new basketball dataset comprising 257 basketball games with 14K event annotations corresponding to 11 event classes. Our model outperforms state-of-the-art methods for both event classification and detection on this new dataset. Additionally, we show that the attention mechanism is able to consistently localize the relevant players.

Nan Ding, Jia Deng, Kevin Murphy et al.

We consider classification problems in which the label space has structure. A common example is hierarchical label spaces, corresponding to the case where one label subsumes another (e.g., animal subsumes dog). But labels can also be mutually exclusive (e.g., dog vs cat) or unrelated (e.g., furry, carnivore). To jointly model hierarchy and exclusion relations, the notion of a HEX (hierarchy and exclusion) graph was introduced in [7]. This combined a conditional random field (CRF) with a deep neural network (DNN), resulting in state of the art results when applied to visual object classification problems where the training labels were drawn from different levels of the ImageNet hierarchy (e.g., an image might be labeled with the basic level category "dog", rather than the more specific label "husky"). In this paper, we extend the HEX model to allow for soft or probabilistic relations between labels, which is useful when there is uncertainty about the relationship between two labels (e.g., an antelope is "sort of" furry, but not to the same degree as a grizzly bear). We call our new model pHEX, for probabilistic HEX. We show that the pHEX graph can be converted to an Ising model, which allows us to use existing off-the-shelf inference methods (in contrast to the HEX method, which needed specialized inference algorithms). Experimental results show significant improvements in a number of large-scale visual object classification tasks, outperforming the previous HEX model.

Maximilian Nickel, Kevin Murphy, Volker Tresp et al.

Relational machine learning studies methods for the statistical analysis of relational, or graph-structured, data. In this paper, we provide a review of how such statistical models can be "trained" on large knowledge graphs, and then used to predict new facts about the world (which is equivalent to predicting new edges in the graph). In particular, we discuss two fundamentally different kinds of statistical relational models, both of which can scale to massive datasets. The first is based on latent feature models such as tensor factorization and multiway neural networks. The second is based on mining observable patterns in the graph. We also show how to combine these latent and observable models to get improved modeling power at decreased computational cost. Finally, we discuss how such statistical models of graphs can be combined with text-based information extraction methods for automatically constructing knowledge graphs from the Web. To this end, we also discuss Google's Knowledge Vault project as an example of such combination.

Tim Althoff, Xin Luna Dong, Kevin Murphy et al.

We present a method called TIMEMACHINE to generate a timeline of events and relations for entities in a knowledge base. For example for an actor, such a timeline should show the most important professional and personal milestones and relationships such as works, awards, collaborations, and family relationships. We develop three orthogonal timeline quality criteria that an ideal timeline should satisfy: (1) it shows events that are relevant to the entity; (2) it shows events that are temporally diverse, so they distribute along the time axis, avoiding visual crowding and allowing for easy user interaction, such as zooming in and out; and (3) it shows events that are content diverse, so they contain many different types of events (e.g., for an actor, it should show movies and marriages and awards, not just movies). We present an algorithm to generate such timelines for a given time period and screen size, based on submodular optimization and web-co-occurrence statistics with provable performance guarantees. A series of user studies using Mechanical Turk shows that all three quality criteria are crucial to produce quality timelines and that our algorithm significantly outperforms various baseline and state-of-the-art methods.

Xin Luna Dong, Evgeniy Gabrilovich, Kevin Murphy et al.

The quality of web sources has been traditionally evaluated using exogenous signals such as the hyperlink structure of the graph. We propose a new approach that relies on endogenous signals, namely, the correctness of factual information provided by the source. A source that has few false facts is considered to be trustworthy. The facts are automatically extracted from each source by information extraction methods commonly used to construct knowledge bases. We propose a way to distinguish errors made in the extraction process from factual errors in the web source per se, by using joint inference in a novel multi-layer probabilistic model. We call the trustworthiness score we computed Knowledge-Based Trust (KBT). On synthetic data, we show that our method can reliably compute the true trustworthiness levels of the sources. We then apply it to a database of 2.8B facts extracted from the web, and thereby estimate the trustworthiness of 119M webpages. Manual evaluation of a subset of the results confirms the effectiveness of the method.

George Papandreou, Liang-Chieh Chen, Kevin Murphy et al.

Deep convolutional neural networks (DCNNs) trained on a large number of images with strong pixel-level annotations have recently significantly pushed the state-of-art in semantic image segmentation. We study the more challenging problem of learning DCNNs for semantic image segmentation from either (1) weakly annotated training data such as bounding boxes or image-level labels or (2) a combination of few strongly labeled and many weakly labeled images, sourced from one or multiple datasets. We develop Expectation-Maximization (EM) methods for semantic image segmentation model training under these weakly supervised and semi-supervised settings. Extensive experimental evaluation shows that the proposed techniques can learn models delivering competitive results on the challenging PASCAL VOC 2012 image segmentation benchmark, while requiring significantly less annotation effort. We share source code implementing the proposed system at https://bitbucket.org/deeplab/deeplab-public.

Liang-Chieh Chen, George Papandreou, Iasonas Kokkinos et al.

Deep Convolutional Neural Networks (DCNNs) have recently shown state of the art performance in high level vision tasks, such as image classification and object detection. This work brings together methods from DCNNs and probabilistic graphical models for addressing the task of pixel-level classification (also called "semantic image segmentation"). We show that responses at the final layer of DCNNs are not sufficiently localized for accurate object segmentation. This is due to the very invariance properties that make DCNNs good for high level tasks. We overcome this poor localization property of deep networks by combining the responses at the final DCNN layer with a fully connected Conditional Random Field (CRF). Qualitatively, our "DeepLab" system is able to localize segment boundaries at a level of accuracy which is beyond previous methods. Quantitatively, our method sets the new state-of-art at the PASCAL VOC-2012 semantic image segmentation task, reaching 71.6% IOU accuracy in the test set. We show how these results can be obtained efficiently: Careful network re-purposing and a novel application of the 'hole' algorithm from the wavelet community allow dense computation of neural net responses at 8 frames per second on a modern GPU.