United States

The increasing of pre-trained models has significantly facilitated the performance on limited data tasks with transfer learning. However, progress on transfer learning mainly focuses on optimizing the weights of pre-trained models, which ignores the structure mismatch between the model and the target task. This paper aims to improve the transfer performance from another angle - in addition to tuning the weights, we tune the structure of pre-trained models, in order to better match the target task. To this end, we propose TransTailor, targeting at pruning the pre-trained model for improved transfer learning. Different from traditional pruning pipelines, we prune and fine-tune the pre-trained model according to the target-aware weight importance, generating an optimal sub-model tailored for a specific target task. In this way, we transfer a more suitable sub-structure that can be applied during fine-tuning to benefit the final performance. Extensive experiments on multiple pre-trained models and datasets demonstrate that TransTailor outperforms the traditional pruning methods and achieves competitive or even better performance than other state-of-the-art transfer learning methods while using a smaller model. Notably, on the Stanford Dogs dataset, TransTailor can achieve 2.7% accuracy improvement over other transfer methods with 20% fewer FLOPs.

AAAI 2021

TransTailor: Pruning the Pre-Trained Model for Improved Transfer Learning

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. 

AAAI 2022

technical paper

In this paper, we study the problem of learning representations of entities and relations in the knowledge graph for the link prediction task. Our idea is based on the observation that the vast majority of the related work only models the relation as a single geometric operation such as translation or rotation, which limits the representation power of the underlying models and makes it harder to match the complicated relations existed in real-world datasets. To embrace a richer set of relational information, we propose a new method called dual quaternion knowledge graph embedding (DualE), which introduces dual quaternions into knowledge graph embeddings. Specifically, a dual quaternion behaves like a â€œcomplex quaternionâ€ with its real and imaginary part all being quaternary. The core of DualE lies a specific design of dual-quaternion-based multiplication, which universally models relations as the compositions of a series of translation and rotation operations. The major merits of DualE are three-fold:1) it is the first unified framework embracing both rotation based and translation-based models, 2) it expands the embedding space to the dual quaternion space with a more intuitive physical and geometric interpretation, 3) it  satisfies the key patterns and the multiple relations pattern of relational representation learning. Experimental results on four real-world datasets demonstrate the effectiveness of our DualE method.

Dual Quaternion Knowledge Graph Embeddings

In classic network security games, the defender distributes defending resources to the nodes of the network, and the attacker attacks a node, with the objective to maximize the damage caused. Existing models assume that the attack at node u causes damage only at u. However, in many real-world security scenarios, the attack at a node u spreads to the neighbors of u and can cause damage at multiple nodes, e.g., for the outbreak of a virus. In this paper, we consider the network defending problem against contagious attacks.
Existing works that study shared resources assume that the resource allocated to a node can be shared or duplicated between neighboring nodes. However, in real world, sharing resource naturally leads to a decrease in defending power of the source node, especially when defending against contagious attacks. To this end, we study the model in which resources allocated to a node can only be transferred to its neighboring nodes, which we refer to as a reallocation process.
We show that this more general model is difficult in two aspects: (1) even for a fixed allocation of resources, we show that computing the optimal reallocation is NP-hard; (2) for the case when reallocation is not allowed, we show that computing the optimal allocation (against contagious attack) is also NP-hard. For positive results, we give a mixed integer linear program formulation for the problem and a bi-criteria approximation algorithm. Our experimental results demonstrate that the allocation and reallocation strategies our algorithm computes perform well in terms of minimizing the damage due to contagious attacks.

Defending against Contagious Attacks on a Network with Resource Reallocation

Machine Learning models should ideally be compact and robust. Compactness provides efficiency and comprehensibility whereas robustness provides stability. Both topics have been studied in recent years but in isolation. Here we present a robust model compression scheme which is independent of model types: it can compress ensembles, neural networks and other types of models into diverse types of small models. 
The main building block is the notion of depth derived from robust statistics. 
Originally, depth was introduced as a measure of the centrality of a point in a sample such that the median is the deepest point.
This concept was extended to classification functions which makes it possible to define the depth of a hypothesis and the median hypothesis. Algorithms have been suggested to approximate the median but they have been limited to binary classification. In this study, we present a new algorithm, the Multiclass Empirical Median Optimization (MEMO) algorithm that finds a deep hypothesis in multi-class tasks, and prove its correctness.
This led to our Compact Robust Estimated Median Belief Optimization (CREMBO) algorithm for robust model compression. We demonstrate the success of this algorithm empirically by compressing neural networks and random forests into small decision trees, which are interpretable models, and show that they are more accurate and robust than other comparable methods. In addition, our empirical study shows that our method outperforms Knowledge Distillation on DNN to DNN compression.

Robust Model Compression Using Deep Hypotheses

Distance Metric Learning (DML) seeks to learn a discriminative embedding where similar examples are closer, and dissimilar examples are apart. In this paper, we address the problem of Semi-Supervised DML (SSDML) that tries to learn a metric using a few labeled examples, and abundantly available unlabeled examples. SSDML is important because it is infeasible to manually annotate all the examples present in a large dataset. Surprisingly, with the exception of a few classical approaches that learn a linear Mahalanobis metric, SSDML has not been studied in the recent years, and lacks approaches in the deep SSDML scenario. In this paper, we address this challenging problem, and revamp SSDML with respect to deep learning. In particular, we propose a stochastic, graph-based approach that first propagates the affinities between the pairs of examples from labeled data, to that of the unlabeled pairs. The propagated affinities are used to mine triplet based constraints for metric learning. We impose orthogonality constraint on the metric parameters, as it leads to a better performance by avoiding a model collapse.

Semi-Supervised Metric Learning: A Deep Resurrection

It is encouraged to see that progress has been made to bridge videos and natural language. However, mainstream video captioning methods suffer from slow inference speed due to the sequential manner of autoregressive decoding, and prefer generating generic descriptions due to the insufficient training of visual words (e.g., nouns and verbs) and inadequate decoding paradigm. In this paper, we propose a non-autoregressive decoding based model with a coarse-to-fine captioning procedure to alleviate these defects. In implementations, we employ a bi-directional self-attention based network as our language model for achieving inference speedup, based on which we decompose the captioning procedure into two stages, where the model has different focuses. Specifically, given that visual words determine the semantic correctness of captions, we design a mechanism of generating visual words to not only promote the training of scene-related words but also capture relevant details from videos to construct a coarse-grained sentence ``template''. Thereafter, we devise dedicated decoding algorithms that fill in the ``template'' with suitable words and modify inappropriate phrasing via iterative refinement to obtain a fine-grained description. Extensive experiments on two mainstream video captioning benchmarks, i.e., MSVD and MSR-VTT, demonstrate that our approach achieves state-of-the-art performance, generates diverse descriptions, and obtains high inference efficiency.

Non-Autoregressive Coarse-to-Fine Video Captioning

Self-attention networks have shown remarkable progress in computer vision tasks
such as image classification. The main benefit of the self-attention mechanism is
the ability to capture long-range feature interactions in attention-maps. However,
the computation of attention-maps requires a learnable key, query, and positional
encoding, whose usage is often not intuitive and computationally expensive. To
mitigate this problem, we propose a novel self-attention module with explicitly
modeled attention-maps using only a single learnable parameter for low computational overhead. The design of explicitly modeled attention-maps using geometric prior is based on the observation that the spatial context for a given pixel within an image is mostly dominated by its neighbors, while more distant pixels have a minor contribution. Concretely, the attention-maps are parametrized via simple functions (e.g., Gaussian kernel) with a learnable radius, which is modeled independently of the input content. Our evaluation shows that our method achieves an accuracy improvement of up to 2.2% over the ResNet-baselines in ImageNet ILSVRC and outperforms other self-attention methods such as AA-ResNet152 in accuracy by 0.9% with 6.4% fewer parameters and 6.7% fewer GFLOPs. This result empirically indicates the value of incorporating geometric prior into self-attention mechanism when applied in image classification.

Explicitly Modeled Attention Maps for Image Classification

We study a model of proxy voting where the candidates, voters, and proxies are all located on the real line, and instead of voting directly, each voter delegates its vote to the closest proxy. The goal is to find a set of proxies that is theta-representative, which entails that for any voter located anywhere on the line, its favorite candidate is within a distance theta of the favorite candidate of its closest proxy. This property guarantees a strong form of representation as the set of voters is not required to be fixed in advance, or even be finite. We show that for candidates located on a line, an optimal proxy arrangement can be computed in polynomial time. Moreover, we provide upper and lower bounds on the number of proxies required to form a theta-representative set, thus showing that a relatively small number of proxies is enough to capture the preferences of any set of voters. An additional beneficial property of a theta-representative proxy arrangement is that for strict-Condorcet voting rules, the outcome of proxy voting is similarly close to the outcome of direct voting.

Representative Proxy Voting

While Machine Comprehension (MC) has attracted extensive research interests in recent years, existing approaches mainly belong to the category of Machine Reading Comprehension task which mines textual inputs (paragraphs and questions) to predict the answers (choices or text spans). However, there are a lot of MC tasks that accept audio input in addition to the textual input, e.g. English listening comprehension test. In this paper, we target the problem of Audio-Oriented Multimodal Machine Comprehension, and its goal is to answer questions based on the given audio and textual information. To solve this problem, we propose a Dynamic Inter- and Intra-modality Attention (DIIA) model to effectively fuse the two modalities (audio and textual). DIIA can work as an independent component and thus be easily integrated into existing MC models. Moreover, we further develop a Multimodal Knowledge Distillation (MKD) module to enable our multimodal MC model to accurately predict the answers based only on either the text or the audio. As a result, the proposed approach can handle various tasks including: Audio-Oriented Multimodal Machine Comprehension, Machine Reading Comprehension and Machine Listening Comprehension, in a single model, making fair comparisons possible between our model and the existing unimodal MC models. Experimental results and analysis prove the effectiveness of the proposed approaches. First, the proposed DIIA boosts the baseline models by up to 21.08% in terms of accuracy; Second, under the unimodal scenarios, the MKD module allows our multimodal MC model to significantly outperform the unimodal models by up to 18.87%, which are trained and tested with only audio or textual data.

Audio-Oriented Multimodal Machine Comprehension via Dynamic Inter- and Intra-Modality Attention

This paper studies the robustness of reinforcement learning algorithms to errors in the learning process. Specifically, we revisit the benchmark problem of discrete-time linear quadratic regulation (LQR) and study the long-standing open question: Under what conditions is the policy iteration method robustly stable from a dynamical systems perspective? Using advanced stability results in control theory, it is shown that policy iteration for LQR is inherently robust to small errors in the learning process and enjoys small-disturbance input-to-state stability: whenever the error in each iteration is bounded and small, the solutions of the policy iteration algorithm are also bounded, and, moreover, enter and stay in a small neighborhood of the optimal LQR solution. As an application, a novel off-policy optimistic least-squares policy iteration for the LQR problem is proposed, when the system dynamics are subjected to additive stochastic disturbances. The proposed new results in robust reinforcement learning are validated by a numerical example.

Robust Reinforcement Learning: A Case Study in Linear Quadratic Regulation

Image matting is a key technique for image and video editing and composition. Conventionally, deep learning approaches take the whole input image and an associated trimap to infer the alpha matte using convolutional neural networks. Such approaches set state-of-the-arts in image matting; however, they may fail in real-world matting applications due to hardware limitations, since real-world input images for matting are mostly of very high resolution. In this paper, we propose HDMatt, a first deep learning based image matting approach for high-resolution inputs. More concretely, HDMatt runs matting in a patch-based crop-and-stitch manner for high-resolution inputs with a novel module design to address the contextual dependency and consistency issues between different patches. Compared with vanilla patch-based inference which computes each patch independently, we explicitly model the cross-patch contextual dependency with a newly-proposed Cross-Patch Contextual module (CPC) guided by the given trimap. Extensive experiments demonstrate the effectiveness of the proposed method and its necessity for high-resolution inputs. Our HDMatt approach also sets new state-of-the-art performance on Adobe Image Matting and AlphaMatting benchmarks and produce impressive visual results on more real-world high-resolution images.

Downloads

Next from AAAI 2021

Dual Quaternion Knowledge Graph Embeddings

Similar lecture

On the Transferability of Pre-Trained Language Models: A Study from Artificial Datasets

.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

Dual Quaternion Knowledge Graph Embeddings

Similar lecture

On the Transferability of Pre-Trained Language Models: A Study from Artificial Datasets

Downloads