United States

An unbiased low-variance gradient estimator, termed GO gradient, was proposed recently for expectation-based objectives E_q_Î³(y) [f(y)], where the random variable (RV) y may be drawn from a stochastic computation graph (SCG) with continuous (non-reparameterizable) internal nodes and continuous/discrete leaves. Based on the GO gradient, we present for E_q_Î³(y) [f(y)] an unbiased low-variance Hessian estimator, named GO Hessian, which contains the deterministic Hessian as a special case. Considering practical implementation, we reveal that the GO Hessian in expectation obeys the chain rule and is therefore easy-to-use with auto-differentiation and Hessian-vector products, enabling efficient cheap exploitation of curvature information over deep SCGs. As representative examples, we present the GO Hessian for non-reparameterizable gamma and negative binomial RVs/nodes. Leveraging the GO Hessian, we develop a new second-order method for E_q_Î³(y) [f(y)], with challenging experiments conducted to verify its effectiveness and efficiency.

AAAI 2021

GO Hessian for Expectation-Based Objectives

The purpose of the AAAI conference is to promote research in artificial intelligence (AI) and scientific exchange among AI researchers, practitioners, scientists, and engineers in affiliated disciplines. 

technical paper

Deep convolutional neural networks (CNNs) have shown a strong ability in mining discriminative object pose and parts information for image recognition. For fine-grained recognition, context-aware rich feature representation of object/scene plays a key role since it exhibits a significant variance in the same subcategory and subtle variance among different subcategories. Finding the subtle variance that fully characterizes the object/scene is not straightforward. To address this, we propose a novel context-aware attentional pooling (CAP) that effectively captures subtle changes via sub-pixel gradients, and learns to attend informative integral regions and their importance in discriminating different subcategories without requiring the bounding-box and/or distinguishable part annotations. We also introduce a novel feature encoding by considering the intrinsic consistency between the informativeness of the integral regions and their spatial structures to capture the semantic correlation among them. Our approach is simple yet extremely effective and can be easily applied on top of a standard classification backbone network. We evaluate our approach using six state-of-the-art (SotA) backbone networks and eight benchmark datasets. Our method significantly outperforms the SotA approaches on six datasets and is very competitive with the remaining two.

Context-Aware Attentional Pooling (CAP) for Fine-Grained Visual Classification

Dynamic load balancing lies at the heart of distributed caching. Here, the goal is to assign objects (load) to servers (computing nodes) in a way that provides load balancing while at the same time dynamically adjusts to the addition or removal of servers. Load balancing is a critical topic in many areas including cloud systems, distributed databases, and distributed and data-parallel machine learning. A popular and widely adopted solution to dynamic load balancing is the two-decade-old Consistent Hashing (CH). Recently, an elegant extension was provided to account for server bounds. In this paper, we identify that existing methodologies for CH and its variants suffer from cascaded overflow, leading to poor load balancing. This cascading effect leads to decreasing performance of the hashing procedure with increasing load. To overcome the cascading effect, we propose a simple solution to CH based on recent advances in fast minwise hashing. We show, both theoretically and empirically, that our proposed solution is significantly superior for load balancing and is optimal in many senses. On the AOL search dataset and Indiana University Clicks dataset with real user activity, our proposed solution reduces cache misses by several magnitudes.

Revisiting Consistent Hashing with Bounded Loads

Graphical event models are representations that capture process independence between different types of events in multivariate temporal point processes. The literature consists of various parametric models and approaches to learn them from multivariate event stream data. Since these models are interpretable, they are often able to provide beneficial insights about event dynamics. In this paper, we show how to compactly model the situation where the order of occurrences of an eventâ€™s causes in some recent historical time interval impacts its occurrence rate; this sort of historical dependence is common in several real-world applications. To overcome the practical challenge of parameter explosion due to the number of potential orders that is super-exponential in the number of parents, we introduce a novel graphical event model based on a parametric tree representation for capturing ordinal historical dependence. We present an approach to learn such a model from data, demonstrating that the proposed model fits several real-world datasets better than relevant baselines. We also showcase the potential advantages of such a model to an analyst during the process of knowledge discovery.

Ordinal Historical Dependence in Graphical Event Models with Tree Representations

The prevalent approach to problems of credit assignment in machine learning -- such as feature and data valuation -- is to model the problem at hand as a cooperative game and apply the Shapley value. But cooperative game theory offers a rich menu of alternative solution concepts, which famously includes the core and its variants. Our goal is to challenge the machine learning community's current consensus around the Shapley value, and make a case for the core as a viable alternative. To that end, we prove that arbitrarily good approximations to the least core -- a core relaxation that is always feasible -- can be computed efficiently (but prove an impossibility for a more refined solution concept, the nucleolus). We also perform experiments that corroborate these theoretical results and shed light on settings where the least core may be preferable to the Shapley value.

If You Like Shapley Then You’ll Love the Core

Spiking neural network (SNN) is promising but the development has fallen far behind conventional deep neural networks (DNNs) because of difficult training. To resolve the training problem, we analyze the closed-form input-output response of spiking neurons and use the response expression to build abstract SNN models for training. This avoids calculating membrane potential during training and makes the direct training of SNN as efficient as DNN. We show that the nonleaky integrate-and-fire neuron with single-spike temporal-coding is the best choice for direct-train deep SNNs. We develop an energy-efficient phase-domain signal processing circuit for the neuron and propose a direct-train deep SNN framework. Thanks to easy training, we train deep SNNs under weight quantizations to study their robustness over low-cost neuromorphic hardware.  Experiments show that our direct-train deep SNNs have the highest CIFAR-10 classification accuracy among SNNs, achieve ImageNet classification accuracy within 1% of the DNN of equivalent architecture, and are robust to weight quantization and noise perturbation.

Temporal-Coded Deep Spiking Neural Network with Easy Training and Robust Performance

Transferring knowledge among various environments is important for efficiently learning multiple tasks online. Most existing methods directly use the previously learned models or previously learned optimal policies to learn new tasks. However, these methods may be inefficient when the underlying models or optimal policies are substantially different across tasks. In this paper, we propose Template Learning (TempLe), a PAC-MDP method for multi-task reinforcement learning that could be applied to tasks with varying state/action space without prior knowledge of inter-task mappings. TempLe gains sample efficiency by extracting similarities of the transition dynamics across tasks even when their underlying models or optimal policies have limited commonalities. We present two algorithms for an ``online'' and a ``finite-model'' setting respectively. We prove that our proposed TempLe algorithms achieve much lower sample complexity than single-task learners or state-of-the-art multi-task methods. We show via systematically designed experiments that our TempLe method universally outperforms the state-of-the-art multi-task methods (PAC-MDP or not) in various settings and regimes.

TempLe: Learning Template of Transitions for Sample Efficient Multi-Task RL

Recent advances in Deep Gaussian Processes (DGPs) show the potential to have more expressive representation than that of traditional Gaussian Processes (GPs). However, there exists a pathology of deep Gaussian processes that their learning capacities reduce significantly when the number of layers increases. In this paper, we present a new analysis in DGPs by studying its corresponding nonlinear dynamic systems to explain the issue. Existing work reports the pathology for the squared exponential kernel function. We extend our investigation to four types of common stationary kernel functions. The recurrence relations between layers are analytically derived, providing a tighter bound and the rate of convergence of the dynamic systems. We demonstrate our finding with a number of experimental results.

Characterizing Deep Gaussian Processes via Nonlinear Recurrence Systems

Cross-lingual models trained on source language tasks possess the capability to directly transfer to target languages. However, since word order variances generally exist in different languages, cross-lingual models that overfit into the word order of the source language could have sub-optimal performance in target languages. In this paper, we hypothesize that reducing the word order information fitted into the models can improve the adaptation performance in target languages. To verify this hypothesis, we introduce several methods to make models encode less word order information of the source language and test them based on cross-lingual word embeddings and the pre-trained multilingual model. Experimental results on three sequence labeling tasks (i.e., part-of-speech tagging, named entity recognition and slot filling tasks) show that reducing word order information injected into the model can achieve better zero-shot cross-lingual performance. Further analysis illustrates that fitting excessive or insufficient word order information into the model results in inferior cross-lingual performance. Moreover, our proposed methods can also be applied to strong cross-lingual models and further improve their performance.

On the Importance of Word Order Information in Cross-Lingual Sequence Labeling

Recognizing humor from a video utterance requires understanding the verbal and non-verbal components as well as incorporating the appropriate context and external knowledge. In this paper, we propose Humor Knowledge enriched Transformer (HKT) that can capture the gist of a multimodal humorous expression by integrating the preceding context and external knowledge. We incorporate humor centric external knowledge into the model by capturing the ambiguity and sentiment present in the language. We encode all the language, acoustic, vision, and humor centric features separately using Transformer based encoders, followed by a cross attention layer to exchange information among them. Our model achieves 77.36% and 79.41% accuracy in humorous punchline detection on UR-FUNNY and MUStaRD datasets -- achieving a new state-of-the-art on both datasets with the margin of 4.93% and 2.94% respectively. Furthermore, we demonstrate that our model can capture interpretable, humor-inducing patterns from all modalities.

Humor Knowledge Enriched Transformer for Understanding Multimodal Humor

In this paper, we first study the problem of combinatorial pure exploration with full-bandit feedback (CPE-BL), where a learner is given a combinatorial action space X \subseteq {0,1}^d, and in each round the learner pulls an action x \in X and receives a random reward with expectation x^T \theta, with \theta \in \R^d a latent and unknown environment vector. 
The objective is to identify the optimal action with the highest expected reward, using as few samples as possible. For CPE-BL, we design the first polynomial-time adaptive algorithm, whose sample complexity matches the lower bound (within a logarithmic factor) for a family of instances and has a light dependence of \Delta_min (the smallest gap between the optimal action and sub-optimal actions). Furthermore, we propose a novel generalization of CPE-BL with  flexible feedback structures, called combinatorial pure exploration with partial linear feedback (CPE-PL), which encompasses several families of sub-problems including full-bandit feedback, semi-bandit feedback, partial feedback and nonlinear reward functions. In CPE-PL, each pull of action x reports a random feedback vector with expectation of M_x \theta , where M_x \in R^{m_x \times d} is a transformation matrix for x, and gains a random (possibly nonlinear) reward related to x.  For CPE-PL, we develop the first polynomial-time algorithm, which simultaneously addresses limited feedback, general reward function and combinatorial action space (e.g., matroids, matchings and s-t paths), and provide its sample complexity analysis. Our empirical evaluation demonstrates that our algorithms run orders of magnitude faster than the existing ones, and our CPE-BL algorithm is robust across different \Delta_min settings while our CPE-PL algorithm is the first one returning correct answers for nonlinear reward functions.

Downloads

Next from AAAI 2021

Context-Aware Attentional Pooling (CAP) for Fine-Grained Visual Classification

Similar lecture

A Deeper Look at the Hessian Eigenspectrum of Deep Neural Networks and Its Applications to Regularization

.css-70qvj9{display:-webkit-box;display:-webkit-flex;display:-ms-flexbox;display:flex;-webkit-align-items:center;-webkit-box-align:center;-ms-flex-align:center;align-items:center;}Downloads

Next from AAAI 2021

Context-Aware Attentional Pooling (CAP) for Fine-Grained Visual Classification

Similar lecture

A Deeper Look at the Hessian Eigenspectrum of Deep Neural Networks and Its Applications to Regularization

Downloads