Adversarial Reinforced Instruction Attacker for Robust Vision-Language Navigation

Language instruction plays an essential role in the natural language grounded navigation tasks. However, navigators trained with limited human-annotated instructions may have difficulties in accurately capturing key information from the complicated instruction at different timesteps, leading to poor navigation performance. In this paper, we exploit to train a more robust navigator which is capable of dynamically extracting crucial factors from the long instruction, by using an adversarial attacking paradigm. Specifically, we propose a Dynamic Reinforced Instruction Attacker (DR-Attacker), which learns to mislead the navigator to move to the wrong target by destroying the most instructive information in instructions at different timesteps. By formulating the perturbation generation as a Markov Decision Process, DR-Attacker is optimized by the reinforcement learning algorithm to generate perturbed instructions sequentially during the navigation, according to a learnable attack score. Then, the perturbed instructions, which serve as hard samples, are used for improving the robustness of the navigator with an effective adversarial training strategy and an auxiliary self-supervised reasoning task. Experimental results on both Vision-and-Language Navigation (VLN) and Navigation from Dialog History (NDH) tasks show the superiority of our proposed method over state-of-the-art methods. Moreover, the visualization analysis shows the effectiveness of the proposed DR-Attacker, which can successfully attack crucial information in the instructions at different timesteps. Code is available at https://github.com/expectorlin/DR-Attacker.

Communication-efficient Decentralized Machine Learning over Heterogeneous Networks

In the last few years, distributed machine learning has been usually executed over heterogeneous networks such as a local area network within a multi-tenant cluster or a wide area network connecting data centers and edge clusters. In these heterogeneous networks, the link speeds among worker nodes vary significantly, making it challenging for state-of-the-art machine learning approaches to perform efficient training. Both centralized and decentralized training approaches suffer from low-speed links. In this paper, we propose a decentralized approach, namely NetMax, that enables worker nodes to communicate via high-speed links and, thus, significantly speed up the training process. NetMax possesses the following novel features. First, it consists of a novel consensus algorithm that allows worker nodes to train model copies on their local dataset asynchronously and exchange information via peer-to-peer communication to synchronize their local copies, instead of a central master node (i.e., parameter server). Second, each worker node selects one peer randomly with a fine-tuned probability to exchange information per iteration. In particular, peers with high-speed links are selected with high probability. Third, the probabilities of selecting peers are designed to minimize the total convergence time. Moreover, we mathematically prove the convergence of NetMax. We evaluate NetMax on heterogeneous cluster networks and show that it achieves speedups of 3.7X, 3.4X, and 1.9X in comparison with the state-of-the-art decentralized training approaches Prague, Allreduce-SGD, and AD-PSGD, respectively.

ForkBase: Immutable, Tamper-evident Storage Substrate for Branchable Applications

Data collaboration activities typically require systematic or protocol-based coordination to be scalable. Git, an effective enabler for collaborative coding, has been attested for its success in countless projects around the world. Hence, applying the Git philosophy to general data collaboration beyond coding is motivating. We call it Git for data. However, the original Git design handles data at the file granule, which is considered too coarse-grained for many database applications. We argue that Git for data should be co-designed with database systems. To this end, we developed ForkBase to make Git for data practical. ForkBase is a distributed, immutable storage system designed for data version management and data collaborative operation. In this demonstration, we show how ForkBase can greatly facilitate collaborative data management and how its novel data deduplication technique can improve storage efficiency for archiving massive da

A single photonic cavity with two independent physical synthetic dimensions

The concept of synthetic dimensions, which has enabled the study of higher-dimensional physics on lower-dimensional physical structures, has generated significant recent interest in many branches of science ranging from ultracold-atomic physics to photonics, since such a concept provides a versatile platform for realizing effective gauge potentials and novel topological physics. Previous experiments demonstrating this concept have augmented the real-space dimensionality by one additional physical synthetic dimension. Here we endow a single ring resonator with two independent physical synthetic dimensions. Our system consists of a temporally modulated ring resonator with spatial coupling between the clockwise and counterclockwise modes, creating a synthetic Hall ladder along the frequency and pseudospin degrees of freedom for photons propagating in the ring. We experimentally observe a wide variety of rich physics, including effective spin-orbit coupling, magnetic fields, spin-momentum locking, a Meissner-to-vortex phase transition, and chiral currents, completely in synthetic dimensions. Our experiments demonstrate that higher-dimensional physics can be studied in simple systems by leveraging the concept of multiple simultaneous synthetic dimensions.

Experimental demonstration of dynamical input isolation in nonadiabatically modulated photonic cavities

Modulated optical cavities have been proposed and demonstrated for applications in communications, laser frequency stabilization, microwave-to-optical conversion and frequency comb generation. However, most studies are restricted to the adiabatic regime, where either the maximum excursion of the modulation or the modulation frequency itself is below the linewidth of the cavity. Here, using a fiber ring resonator with an embedded electro-optic phase modulator, we investigate the nonadiabatic regime. By strongly driving the modulator at frequencies that are significantly smaller than the free-spectral range of the ring resonator, but well beyond the linewidth of the resonator, we experimentally observe counterintuitive behavior predicted in a recent theoretical study by Minkov et al. [APL Photonics 2, 076101 (2017)], such as the complete suppression of drop-port transmission even when the input laser wavelength is on resonance with the optical cavity. This can be understood as dynamical isolation of the cavity from the input light. We also show qualitative differences in the steady-state responses of the system between the adiabatic and nonadiabatic limits. Our experiments probe a seldom explored regime of operation that is promising for applications in integrated photonic systems with current state-of-the-art technology.

Jointly Deep Multi-View Learning for Clustering Analysis

In this paper, we propose a novel Joint framework for Deep Multi-view Clustering (DMJC), where multiple deep embedded features, multi-view fusion mechanism and clustering assignments can be learned simultaneously. Our key idea is that the joint learning strategy can sufficiently exploit clustering-friendly multi-view features and useful multi-view complementary information to improve the clustering performance. How to realize the multi-view fusion in such a joint framework is the primary challenge. To do so, we design two ingenious variants of deep multi-view joint clustering models under the proposed framework, where multi-view fusion is implemented by two different schemes. The first model, called DMJC-S, performs multi-view fusion in an implicit way via a novel multi-view soft assignment distribution. The second model, termed DMJC-T, defines a novel multi-view auxiliary target distribution to conduct the multi-view fusion explicitly. Both DMJC-S and DMJC-T are optimized under a KL divergence like clustering objective. Experiments on six challenging image datasets demonstrate the superiority of both DMJC-S and DMJC-T over single/multi-view baselines and the state-of-the-art multiview clustering methods, which proves the effectiveness of the proposed DMJC framework. To our best knowledge, this is the first work to model the multi-view clustering in a deep joint framework, which will provide a meaningful thinking in unsupervised multi-view learning.

ForkBase: An Efficient Storage Engine for Blockchain and Forkable Applications

Existing data storage systems offer a wide range of functionalities to accommodate an equally diverse range of applications. However, new classes of applications have emerged, e.g., blockchain and collaborative analytics, featuring data versioning, fork semantics, tamper-evidence or any combination thereof. They present new opportunities for storage systems to efficiently support such applications by embedding the above requirements into the storage. In this paper, we present ForkBase, a storage engine specifically designed to provide efficient support for blockchain and forkable applications. By integrating the core application properties into the storage, ForkBase not only delivers high performance but also reduces development effort. Data in ForkBase is multi-versioned, and each version uniquely identifies the data content and its history. Two variants of fork semantics are supported in ForkBase to facilitate any collaboration workflows. A novel index structure is introduced to efficiently identify and eliminate duplicate content across data objects. Consequently, ForkBase is not only efficient in performance, but also in space requirement. We demonstrate the performance of ForkBase using three applications: a blockchain platform, a wiki engine and a collaborative analytics application. We conduct extensive experimental evaluation of these applications against respective state-of-the-art system. The results show that ForkBase achieves superior performance while significantly lowering the development cost.

Constructing three-dimensional photonic topological insulator using two-dimensional ring resonator lattice with a synthetic frequency dimension

In the development of topological photonics, achieving three dimensional topological insulators is of significant interest since it enables the exploration of new topological physics with photons, and promises novel photonic devices that are robust against disorders in three dimensions. Previous theoretical proposals towards three dimensional topological insulators utilize complex geometries that are challenging to implement. Here, based on the concept of synthetic dimension, we show that a two-dimensional array of ring resonators, which was previously demonstrated to exhibit a two-dimensional topological insulator phase, in fact automatically becomes a three-dimensional topological insulator, when the frequency dimension is taken into account. Moreover, by modulating a few of the resonators, a screw dislocation along the frequency axis can be created, which provides robust transport of photons along the frequency axis. Demonstrating the physics of screw dislocation in a topological system has been a significant challenge in solid state systems. Our work indicates that the physics of three-dimensional topological insulator can be explored in standard integrated photonics platforms, leading to opportunities for novel devices that control the frequency of light.

Synthetic space with arbitrary dimensions in a few rings undergoing dynamic modulation

We show that a single ring resonator undergoing dynamic modulation can be used to create a synthetic space with an arbitrary dimension. In such a system the phases of the modulation can be used to create a photonic gauge potential in high dimensions. As an illustration of the implication of this concept, we show that the Haldane model, which exhibits non-trivial topology in two dimensions, can be implemented in the synthetic space using three rings. Our results point to a route towards exploring higher-dimensional topological physics in low-dimensional physical structures. The dynamics of photons in such synthetic spaces also provides a mechanism to control the spectrum of light.

Scaling Distributed Transaction Processing and Recovery based on Dependency Logging

DGCC protocol has been shown to achieve good performance on multi-core in-memory system. However, distributed transactions complicate the dependency resolution, and therefore, an effective transaction partitioning strategy is essential to reduce expensive multi-node distributed transactions. During failure recovery, log must be examined from the last checkpoint onwards and the affected transactions are re-executed based on the way they are partitioned and executed. Existing approaches treat both transaction management and recovery as two separate problems, even though recovery is dependent on the sequence in which transactions are executed. In this paper, we propose to treat the transaction management and recovery problems as one. We first propose an efficient Distributed Dependency Graph based Concurrency Control (DistDGCC) protocol for handling transactions spanning multiple nodes, and propose a new novel and efficient logging protocol called Dependency Logging that also makes use of dependency graphs for efficient logging and recovery. DistDGCC optimizes the average cost for each distributed transaction by processing transactions in batch. Moreover, it also reduces the effects of thread blocking caused by distributed transactions and consequently improves the runtime performance. Further, dependency logging exploits the same data structure that is used by DistDGCC to reduce the logging overhead, as well as the logical dependency information to improve the recovery parallelism. Extensive experiments are conducted to evaluate the performance of our proposed technique against state-of-the-art techniques. Experimental results show that DistDGCC is efficient and scalable, and dependency logging supports fast recovery with marginal runtime overhead. Hence, the overall system performance is significantly improved as a result.