With the rapid growth of deep learning in many fields, machine learning-assisted communication systems had attracted lots of researches with many eye-catching initial results. At the present stage, most of the methods still have great demand of massive labeled data for supervised learning. However, obtaining labeled data in the practical applications is not feasible, which may result in severe performance degradation due to channel variations. To overcome such a constraint, syndrome loss has been proposed to penalize non-valid decoded codewords and achieve unsupervised learning for neural network-based decoder. However, it cannot be applied to polar decoder directly. In this work, by exploiting the nature of polar codes, we propose a modified syndrome loss. From simulation results, the proposed method demonstrates that domain-specific knowledge and know-how in code structure can enable unsupervised learning for neural network-based polar decoder.
In this paper, we study a simple algorithm to construct asymptotically valid confidence regions for model parameters using the batch means method. The main idea is to cancel out the covariance matrix which is hard/costly to estimate. In the process of developing the algorithm, we establish process-level functional central limit theorem for Polyak-Ruppert averaging based stochastic gradient descent estimators. We also extend the batch means method to accommodate more general batch size specifications.
In the recent past, the success of Neural Architecture Search (NAS) has enabled researchers to broadly explore the design space using learning-based methods. Apart from finding better neural network architectures, the idea of automation has also inspired to improve their implementations on hardware. While some practices of hardware machine-learning automation have achieved remarkable performance, the traditional design concept is still followed: a network architecture is first structured with excellent test accuracy, and then compressed and optimized to fit into a target platform. Such a design flow will easily lead to inferior local-optimal solutions. To address this problem, we propose a new framework to jointly explore the space of neural architecture, hardware implementation, and quantization. Our objective is to find a quantized architecture with the highest accuracy that is implementable on given hardware specifications. We employ FPGAs to implement and test our designs with limited loop-up tables (LUTs) and required throughput. Compared to the separate design/searching methods, our framework has demonstrated much better performance under strict specifications and generated designs of higher accuracy by 18\% to 68\% in the task of classifying CIFAR10 images. With 30,000 LUTs, a light-weight design is found to achieve 82.98\% accuracy and 1293 images/second throughput, compared to which, under the same constraints, the traditional method even fails to find a valid solution.
Composing knowledge from multiple pieces of texts is a key challenge in multi-hop question answering. We present a multi-hop reasoning dataset, Question Answering via Sentence Composition(QASC), that requires retrieving facts from a large corpus and composing them to answer a multiple-choice question. QASC is the first dataset to offer two desirable properties: (a) the facts to be composed are annotated in a large corpus, and (b) the decomposition into these facts is not evident from the question itself. The latter makes retrieval challenging as the system must introduce new concepts or relations in order to discover potential decompositions. Further, the reasoning model must then learn to identify valid compositions of these retrieved facts using common-sense reasoning. To help address these challenges, we provide annotation for supporting facts as well as their composition. Guided by these annotations, we present a two-step approach to mitigate the retrieval challenges. We use other multiple-choice datasets as additional training data to strengthen the reasoning model. Our proposed approach improves over current state-of-the-art language models by 11% (absolute). The reasoning and retrieval problems, however, remain unsolved as this model still lags by 20% behind human performance.
To achieve a successful grasp, gripper attributes such as its geometry and kinematics play a role as important as the object geometry. The majority of previous work has focused on developing grasp methods that generalize over novel object geometry but are specific to a certain robot hand. We propose UniGrasp, an efficient data-driven grasp synthesis method that considers both the object geometry and gripper attributes as inputs. UniGrasp is based on a novel deep neural network architecture that selects sets of contact points from the input point cloud of the object. The proposed model is trained on a large dataset to produce contact points that are in force closure and reachable by the robot hand. By using contact points as output, we can transfer between a diverse set of multifingered robotic hands. Our model produces over 90% valid contact points in Top10 predictions in simulation and more than 90% successful grasps in real world experiments for various known two-fingered and three-fingered grippers. Our model also achieves 93%, 83% and 90% successful grasps in real world experiments for an unseen two-fingered gripper and two unseen multi-fingered anthropomorphic robotic hands.
Conformal prediction is a popular tool for providing valid prediction sets for classification and regression problems, without relying on any distributional assumptions on the data. While the traditional description of conformal prediction starts with a nonconformity score, we provide an alternate (but equivalent) view that starts with a sequence of nested sets and calibrates them to find a valid prediction set. The nested framework subsumes all nonconformity scores, including recent proposals based on quantile regression and density estimation. While these ideas were originally derived based on sample splitting, our framework seamlessly extends them to other aggregation schemes like cross-conformal, jackknife+ and out-of-bag methods. We use the framework to derive a new algorithm (QOOB, pronounced cube) that combines four ideas: quantile regression, cross-conformalization, ensemble methods and out-of-bag predictions. We develop a computationally efficient implementation of cross-conformal, that is also used by QOOB. In a detailed numerical investigation, QOOB performs either the best or close to the best on all simulated and real datasets. Code for QOOB is available at https://github.com/aigen/QOOB.
Natural question generation (QG) aims to generate questions from a passage and an answer. In this paper, we propose a novel reinforcement learning (RL) based graph-to-sequence (Graph2Seq) model for QG. Our model consists of a Graph2Seq generator where a novel Bidirectional Gated Graph Neural Network is proposed to embed the passage, and a hybrid evaluator with a mixed objective combining both cross-entropy and RL losses to ensure the generation of syntactically and semantically valid text. The proposed model outperforms previous state-of-the-art methods by a large margin on the SQuAD dataset.
This paper proposes SplitSGD, a new dynamic learning rate schedule for stochastic optimization. This method decreases the learning rate for better adaptation to the local geometry of the objective function whenever a stationary phase is detected, that is, the iterates are likely to bounce at around a vicinity of a local minimum. The detection is performed by splitting the single thread into two and using the inner product of the gradients from the two threads as a measure of stationarity. Owing to this simple yet provably valid stationarity detection, SplitSGD is easy-to-implement and essentially does not incur additional computational cost than standard SGD. Through a series of extensive experiments, we show that this method is appropriate for both convex problems and training (non-convex) neural networks, with performance compared favorably to other stochastic optimization methods. Importantly, this method is observed to be very robust with a set of default parameters for a wide range of problems and, moreover, can yield better generalization performance than other adaptive gradient methods such as Adam.
In this work, a novel high-speed single object tracker that is robust against non-semantic distractor exemplars is introduced; dubbed BOBBY2. It incorporates a novel exemplar buffer module that sparsely caches the target's appearance across time, enabling it to adapt to potential target deformation. As for training, an augmented ImageNet-VID dataset was used in conjunction with the one cycle policy, enabling it to reach convergence with less than 2 epoch worth of data. For validation, the model was benchmarked on the GOT-10k dataset and on an additional small, albeit challenging custom UAV dataset collected with the TU-3 UAV. We demonstrate that the exemplar buffer is capable of providing redundancies in case of unintended target drifts, a desirable trait in any middle to long term tracking. Even when the buffer is predominantly filled with distractors instead of valid exemplars, BOBBY2 is capable of maintaining a near-optimal level of accuracy. BOBBY2 manages to achieve a very competitive result on the GOT-10k dataset and to a lesser degree on the challenging custom TU-3 dataset, without fine-tuning, demonstrating its generalizability. In terms of speed, BOBBY2 utilizes a stripped down AlexNet as feature extractor with 63% less parameters than a vanilla AlexNet, thus being able to run at a competitive 85 FPS.
Quantizing large Neural Networks (NN) while maintaining the performance is highly desirable for resource-limited devices due to reduced memory and time complexity. It is usually formulated as a constrained optimization problem and optimized via a modified version of gradient descent. In this work, by interpreting the continuous parameters (unconstrained) as the dual of the quantized ones, we introduce a Mirror Descent (MD) framework for NN quantization. Specifically, we provide conditions on the projections (i.e., mapping from continuous to quantized ones) which would enable us to derive valid mirror maps and in turn the respective MD updates. Furthermore, we present a numerically stable implementation of MD that requires storing an additional set of auxiliary variables (unconstrained), and show that it is strikingly analogous to the Straight Through Estimator (STE) based method which is typically viewed as a "trick" to avoid vanishing gradients issue. Our experiments on CIFAR-10/100, TinyImageNet, and ImageNet classification datasets with VGG-16, ResNet-18, and MobileNetV2 architectures show that our MD variants obtain quantized networks with state-of-the-art performance. Code is available at https://github.com/kartikgupta-at-anu/md-bnn.
Do large datasets provide value to psychologists? Without a systematic methodology for working with such datasets, there is a valid concern that analyses will produce noise artifacts rather than true effects. In this paper, we offer a way to enable researchers to systematically build models and identify novel phenomena in large datasets. One traditional approach is to analyze the residuals of models---the biggest errors they make in predicting the data---to discover what might be missing from those models. However, once a dataset is sufficiently large, machine learning algorithms approximate the true underlying function better than the data, suggesting instead that the predictions of these data-driven models should be used to guide model-building. We call this approach "Scientific Regret Minimization" (SRM) as it focuses on minimizing errors for cases that we know should have been predictable. We demonstrate this methodology on a subset of the Moral Machine dataset, a public collection of roughly forty million moral decisions. Using SRM, we found that incorporating a set of deontological principles that capture dimensions along which groups of agents can vary (e.g. sex and age) improves a computational model of human moral judgment. Furthermore, we were able to identify and independently validate three interesting moral phenomena: criminal dehumanization, age of responsibility, and asymmetric notions of responsibility.
Advances in machine intelligence have enabled conversational interfaces that have the potential to radically change the way humans interact with machines. However, even with the progress in the abilities of these agents, there remain critical gaps in their capacity for natural interactions. One limitation is that the agents are often monotonic in behavior and do not adapt to their partner. We built two end-to-end conversational agents: a voice-based agent that can engage in naturalistic, multi-turn dialogue and align with the interlocutor's conversational style, and a 2nd, expressive, embodied conversational agent (ECA) that can recognize human behavior during open-ended conversations and automatically align its responses to the visual and conversational style of the other party. The embodied conversational agent leverages multimodal inputs to produce rich and perceptually valid vocal and facial responses (e.g., lip syncing and expressions) during the conversation. Based on empirical results from a set of user studies, we highlight several significant challenges in building such systems and provide design guidelines for multi-turn dialogue interactions using style adaptation for future research.
Markov chain Monte Carlo (MCMC) methods have not been broadly adopted in Bayesian neural networks (BNNs). This paper initially reviews the main challenges in sampling from the parameter posterior of a neural network via MCMC. Such challenges culminate to lack of convergence to the parameter posterior. Nevertheless, this paper shows that a non-converged Markov chain, generated via MCMC sampling from the parameter space of a neural network, can yield via Bayesian marginalization a valuable posterior predictive distribution of the output of the neural network. Classification examples based on multilayer perceptrons showcase highly accurate posterior predictive distributions. The postulate of limited scope for MCMC developments in BNNs is partially valid; an asymptotically exact parameter posterior seems less plausible, yet an accurate posterior predictive distribution is a tenable research avenue.
We investigate statistical uncertainty quantification for reinforcement learning (RL) and its implications in exploration policy. Despite ever-growing literature on RL applications, fundamental questions about inference and error quantification, such as large-sample behaviors, appear to remain quite open. In this paper, we fill in the literature gap by studying the central limit theorem behaviors of estimated Q-values and value functions under various RL settings. In particular, we explicitly identify closed-form expressions of the asymptotic variances, which allow us to efficiently construct asymptotically valid confidence regions for key RL quantities. Furthermore, we utilize these asymptotic expressions to design an effective exploration strategy, which we call Q-value-based Optimal Computing Budget Allocation (Q-OCBA). The policy relies on maximizing the relative discrepancies among the Q-value estimates. Numerical experiments show superior performances of our exploration strategy than other benchmark policies.
(This comment has been updated to respond to Wang and Blei's rejoinder [arXiv:1910.07320].) The premise of the deconfounder method proposed in "Blessings of Multiple Causes" by Wang and Blei [arXiv:1805.06826], namely that a variable that renders multiple causes conditionally independent also controls for unmeasured multi-cause confounding, is incorrect. This can be seen by noting that no fact about the observed data alone can be informative about ignorability, since ignorability is compatible with any observed data distribution. Methods to control for unmeasured confounding may be valid with additional assumptions in specific settings, but they cannot, in general, provide a checkable approach to causal inference, and they do not, in general, require weaker assumptions than the assumptions that are commonly used for causal inference. While this is outside the scope of this comment, we note that much recent work on applying ideas from latent variable modeling to causal inference problems suffers from similar issues.
Deep learning methods for image quality assessment (IQA) are limited due to the small size of existing datasets. Extensive datasets require substantial resources both for generating publishable content and annotating it accurately. We present a systematic and scalable approach to creating KonIQ-10k, the largest IQA dataset to date, consisting of 10,073 quality scored images. It is the first in-the-wild database aiming for ecological validity, concerning the authenticity of distortions, the diversity of content, and quality-related indicators. Through the use of crowdsourcing, we obtained 1.2 million reliable quality ratings from 1,459 crowd workers, paving the way for more general IQA models. We propose a novel, deep learning model (KonCept512), to show an excellent generalization beyond the test set (0.921 SROCC), to the current state-of-the-art database LIVE-in-the-Wild (0.825 SROCC). The model derives its core performance from the InceptionResNet architecture, being trained at a higher resolution than previous models (512x384). Correlation analysis shows that KonCept512 performs similar to having 9 subjective scores for each test image.
This paper presents an approach to address data scarcity problems in underwater image datasets for visual detection of marine debris. The proposed approach relies on a two-stage variational autoencoder (VAE) and a binary classifier to evaluate the generated imagery for quality and realism. From the images generated by the two-stage VAE, the binary classifier selects "good quality" images and augments the given dataset with them. Lastly, a multi-class classifier is used to evaluate the impact of the augmentation process by measuring the accuracy of an object detector trained on combinations of real and generated trash images. Our results show that the classifier trained with the augmented data outperforms the one trained only with the real data. This approach will not only be valid for the underwater trash classification problem presented in this paper, but it will also be useful for any data-dependent task for which collecting more images is challenging or infeasible.
Generalized planning is the task of generating a single solution that is valid for a set of planning problems. In this paper we show how to represent and compute generalized plans using procedural Domain Control Knowledge (DCK). We define a {\it divide and conquer} approach that first generates the procedural DCK solving a set of planning problems representative of certain subtasks and then compile it as callable procedures of the overall generalized planning problem. Our procedure calling mechanism allows nested and recursive procedure calls and is implemented in PDDL so that classical planners can compute and exploit procedural DCK. Experiments show that an off-the-shelf classical planner, using procedural DCK as callable procedures, can compute generalized plans in a wide range of domains including non-trivial ones, such as sorting variable-size lists or DFS traversal of binary trees with variable size.
Neural backdoor attack is emerging as a severe security threat to deep learning, while the capability of existing defense methods is limited, especially for complex backdoor triggers. In the work, we explore the space formed by the pixel values of all possible backdoor triggers. An original trigger used by an attacker to build the backdoored model represents only a point in the space. It then will be generalized into a distribution of valid triggers, all of which can influence the backdoored model. Thus, previous methods that model only one point of the trigger distribution is not sufficient. Getting the entire trigger distribution, e.g., via generative modeling, is a key to effective defense. However, existing generative modeling techniques for image generation are not applicable to the backdoor scenario as the trigger distribution is completely unknown. In this work, we propose max-entropy staircase approximator (MESA), an algorithm for high-dimensional sampling-free generative modeling and use it to recover the trigger distribution. We also develop a defense technique to remove the triggers from the backdoored model. Our experiments on Cifar10/100 dataset demonstrate the effectiveness of MESA in modeling the trigger distribution and the robustness of the proposed defense method.
When robots work in a cluttered environment, the constraints for motions change frequently and the required action can change even for the same task. However, planning complex motions from direct calculation has the risk of resulting in poor performance local optima. In addition, machine learning approaches often require relearning for novel situations. In this paper, we propose a method of searching appropriate motions by using conditional Generative Adversarial Networks (cGANs), which can generate motions based on the conditions by mimicking training datasets. By training cGANs with various motions for a task, its latent space is fulfilled with the valid motions for the task. The appropriate motions can be found efficiently by searching the latent space of the trained cGANs instead of the motion space, while avoiding poor local optima. We demonstrate that the proposed method successfully works for an object-throwing task to given target positions in both numerical simulation and real-robot experiments. The proposed method resulted in three times higher accuracy with 2.5 times faster calculation time than searching the action space directly.