For a target task where labeled data is unavailable, domain adaptation can transfer a learner from a different source domain. Previous deep domain adaptation methods mainly learn a global domain shift, i.e., align the global source and target distributions without considering the relationships between two subdomains within the same category of different domains, leading to unsatisfying transfer learning performance without capturing the fine-grained information. Recently, more and more researchers pay attention to Subdomain Adaptation which focuses on accurately aligning the distributions of the relevant subdomains. However, most of them are adversarial methods which contain several loss functions and converge slowly. Based on this, we present Deep Subdomain Adaptation Network (DSAN) which learns a transfer network by aligning the relevant subdomain distributions of domain-specific layer activations across different domains based on a local maximum mean discrepancy (LMMD). Our DSAN is very simple but effective which does not need adversarial training and converges fast. The adaptation can be achieved easily with most feed-forward network models by extending them with LMMD loss, which can be trained efficiently via back-propagation. Experiments demonstrate that DSAN can achieve remarkable results on both object recognition tasks and digit classification tasks. Our code will be available at: https://github.com/easezyc/deep-transfer-learning
The Transformer architecture has become a dominant choice in many domains, such as natural language processing and computer vision. Yet, it has not achieved competitive performance on popular leaderboards of graph-level prediction compared to mainstream GNN variants. Therefore, it remains a mystery how Transformers could perform well for graph representation learning. In this paper, we solve this mystery by presenting Graphormer, which is built upon the standard Transformer architecture, and could attain excellent results on a broad range of graph representation learning tasks, especially on the recent OGB Large-Scale Challenge. Our key insight to utilizing Transformer in the graph is the necessity of effectively encoding the structural information of a graph into the model. To this end, we propose several simple yet effective structural encoding methods to help Graphormer better model graph-structured data. Besides, we mathematically characterize the expressive power of Graphormer and exhibit that with our ways of encoding the structural information of graphs, many popular GNN variants could be covered as the special cases of Graphormer.
Semantic understanding of programs is a fundamental problem for programming language processing (PLP). Recent works that learn representations of code based on pre-training techniques in NLP have pushed the frontiers in this direction. However, the semantics of PL and NL have essential differences. These being ignored, we believe it is difficult to build a model to better understand programs, by either directly applying off-the-shelf NLP pre-training techniques to the source code, or adding features to the model by the heuristic. In fact, the semantics of a program can be rigorously defined by formal semantics in PL theory. For example, the operational semantics, describes the meaning of a valid program as updating the environment (i.e., the memory address-value function) through fundamental operations, such as memory I/O and conditional branching. Inspired by this, we propose a novel program semantics learning paradigm, that the model should learn from information composed of (1) the representations which align well with the fundamental operations in operational semantics, and (2) the information of environment transition, which is indispensable for program understanding. To validate our proposal, we present a hierarchical Transformer-based pre-training model called OSCAR to better facilitate the understanding of programs. OSCAR learns from intermediate representation (IR) and an encoded representation derived from static analysis, which are used for representing the fundamental operations and approximating the environment transitions respectively. OSCAR empirically shows the outstanding capability of program semantics understanding on many practical software engineering tasks.
An important development in deep learning from the earliest MLPs has been a move towards architectures with structural inductive biases which enable the model to keep distinct sources of information and routes of processing well-separated. This structure is linked to the notion of independent mechanisms from the causality literature, in which a mechanism is able to retain the same processing as irrelevant aspects of the world are changed. For example, convnets enable separation over positions, while attention-based architectures (especially Transformers) learn which combination of positions to process dynamically. In this work we explore a way in which the Transformer architecture is deficient: it represents each position with a large monolithic hidden representation and a single set of parameters which are applied over the entire hidden representation. This potentially throws unrelated sources of information together, and limits the Transformer's ability to capture independent mechanisms. To address this, we propose Transformers with Independent Mechanisms (TIM), a new Transformer layer which divides the hidden representation and parameters into multiple mechanisms, which only exchange information through attention. Additionally, we propose a competition mechanism which encourages these mechanisms to specialize over time steps, and thus be more independent. We study TIM on a large-scale BERT model, on the Image Transformer, and on speech enhancement and find evidence for semantically meaningful specialization as well as improved performance.
Improving the efficiency of Transformer-based language pre-training is an important task in NLP, especially for the self-attention module, which is computationally expensive. In this paper, we propose a simple but effective solution, called \emph{LazyFormer}, which computes the self-attention distribution infrequently. LazyFormer composes of multiple lazy blocks, each of which contains multiple Transformer layers. In each lazy block, the self-attention distribution is only computed once in the first layer and then is reused in all upper layers. In this way, the cost of computation could be largely saved. We also provide several training tricks for LazyFormer. Extensive experiments demonstrate the effectiveness of the proposed method.
Many real-world applications use Siamese networks to efficiently match text sequences at scale, which require high-quality sequence encodings. This paper pre-trains language models dedicated to sequence matching in Siamese architectures. We first hypothesize that a representation is better for sequence matching if the entire sequence can be reconstructed from it, which, however, is unlikely to be achieved in standard autoencoders: A strong decoder can rely on its capacity and natural language patterns to reconstruct and bypass the needs of better sequence encodings. Therefore we propose a new self-learning method that pretrains the encoder with a weak decoder, which reconstructs the original sequence from the encoder's [CLS] representations but is restricted in both capacity and attention span. In our experiments on web search and recommendation, the pre-trained SEED-Encoder, "SiamEsE oriented encoder by reconstructing from weak decoder", shows significantly better generalization ability when fine-tuned in Siamese networks, improving overall accuracy and few-shot performances. Our code and models will be released.
Transformer has demonstrated its great power to learn contextual word representations for multiple languages in a single model. To process multilingual sentences in the model, a learnable vector is usually assigned to each language, which is called "language embedding". The language embedding can be either added to the word embedding or attached at the beginning of the sentence. It serves as a language-specific signal for the Transformer to capture contextual representations across languages. In this paper, we revisit the use of language embedding and identify several problems in the existing formulations. By investigating the interaction between language embedding and word embedding in the self-attention module, we find that the current methods cannot reflect the language-specific word correlation well. Given these findings, we propose a new approach called Cross-lingual Language Projection (XLP) to replace language embedding. For a sentence, XLP projects the word embeddings into language-specific semantic space, and then the projected embeddings will be fed into the Transformer model to process with their language-specific meanings. In such a way, XLP achieves the purpose of appropriately encoding "language" in a multilingual Transformer model. Experimental results show that XLP can freely and significantly boost the model performance on extensive multilingual benchmark datasets. Codes and models will be released at https://github.com/lsj2408/XLP.
Self-supervised learning, a.k.a., pretraining, is important in natural language processing. Most of the pretraining methods first randomly mask some positions in a sentence and then train a model to recover the tokens at the masked positions. In such a way, the model can be trained without human labeling, and the massive data can be used with billion parameters. Therefore, the optimization efficiency becomes critical. In this paper, we tackle the problem from the view of gradient variance reduction. In particular, we first propose a principled gradient variance decomposition theorem, which shows that the variance of the stochastic gradient of the language pretraining can be naturally decomposed into two terms: the variance that arises from the sample of data in a batch, and the variance that arises from the sampling of the mask. The second term is the key difference between selfsupervised learning and supervised learning, which makes the pretraining slower. In order to reduce the variance of the second part, we leverage the importance sampling strategy, which aims at sampling the masks according to a proposal distribution instead of the uniform distribution. It can be shown that if the proposal distribution is proportional to the gradient norm, the variance of the sampling is reduced. To improve efficiency, we introduced a MAsk Proposal Network (MAPNet), which approximates the optimal mask proposal distribution and is trained end-to-end along with the model. According to the experimental result, our model converges much faster and achieves higher performance than the baseline BERT model.
How to make unsupervised language pre-training more efficient and less resource-intensive is an important research direction in NLP. In this paper, we focus on improving the efficiency of language pre-training methods through providing better data utilization. It is well-known that in language data corpus, words follow a heavy-tail distribution. A large proportion of words appear only very few times and the embeddings of rare words are usually poorly optimized. We argue that such embeddings carry inadequate semantic signals. They could make the data utilization inefficient and slow down the pre-training of the entire model. To solve this problem, we propose Taking Notes on the Fly (TNF). TNF takes notes for rare words on the fly during pre-training to help the model understand them when they occur next time. Specifically, TNF maintains a note dictionary and saves a rare word's context information in it as notes when the rare word occurs in a sentence. When the same rare word occurs again in training, TNF employs the note information saved beforehand to enhance the semantics of the current sentence. By doing so, TNF provides a better data utilization since cross-sentence information is employed to cover the inadequate semantics caused by rare words in the sentences. Experimental results show that TNF significantly expedite the BERT pre-training and improve the model's performance on downstream tasks. TNF's training time is $60\%$ less than BERT when reaching the same performance. When trained with same number of iterations, TNF significantly outperforms BERT on most of downstream tasks and the average GLUE score.