As consensus protocols scale, storage has become the dominant bottleneck in modern blockchains. Systems must maintain historical versions, generate integrity proofs, and sustain high throughput as state grows. Prior work often redesigns the authenticated data structure (ADS), but such changes sacrifice compatibility and require disruptive hard forks. We take the complementary approach: redesigning the storage layer while preserving the existing ADS interface.

Synchronous solutions for Byzantine Fault Tolerance (BFT) can tolerate up to minority faults. In this work, we present Sync HotStuff, a surprisingly simple and intuitive synchronous BFT solution that achieves consensus with a latency of 2Δ in the steady state (where Δ is a synchronous message delay upper bound). In addition, Sync HotStuff ensures safety in a weaker synchronous model in which the synchrony assumption does not have to hold for all replicas all the time. Moreover, Sync HotStuff has optimistic responsiveness, i.e., it advances at network speed when less than one-quarter of the replicas are not responding. Borrowing from practical partially synchronous BFT solutions, Sync HotStuff has a two-phase leader-based structure, and has been fully prototyped under the standard synchrony assumption. When tolerating a single fault, Sync HotStuff achieves a throughput of over 280 Kops/sec under typical network performance, which is comparable to the best known partially synchronous solution.

We present HotStuff, a leader-based Byzantine fault-tolerant replication protocol for the partially synchronous model. Once network communication becomes synchronous, HotStuff enables a correct leader to drive the protocol to consensus at the pace of actual (vs. maximum) network delay--a property called responsiveness---and with communication complexity that is linear in the number of replicas. To our knowledge, HotStuff is the first partially synchronous BFT replication protocol exhibiting these combined properties. Its simplicity enables it to be further pipelined and simplified into a practical, concise protocol for building large-scale replication services.

A recent trend in normalization of factors extraneous to a speech recognition task has been to explicitly introduce features related to the unwanted variability in the training of Deep Neural Networks (DNN). Typically, this is done by either perturbing the training set with models of these extraneous factors such as vocal tract length and environmental noise or augmenting the conventional spectral features with auxiliary information such as i-vector, noise spectrum, etc. Another emerging approach is to derive low dimensional representations of the factors from the hidden layers of DNN and use it for normalization of the acoustic model. Almost all of these approaches focus on either speaker or environment normalization. In this paper we propose a novel approach for estimating a compact joint representation of speakers and environment by training a DNN, with a bottleneck layer, to classify the i-vector features into speaker and environment labels by Multi-Task Learning (MTL). Another novelty is to learn this compact representation while learning to map the i-vector of a noisy utterance into its corresponding clean speaker i-vector and noise-only i-vector. Experiments were conducted on an artificially noise-corrupted version of the WSJ corpus. The proposed compact joint speaker-environment representations show promising gains.

Although deep neural networks (DNNs) have achieved great success in automatic speech recognition (ASR), significant performance degradation still exists in noisy environments. In this paper, a novel multi-task joint-learning framework is proposed to address the noise robustness for speech recognition. The architecture integrates two different DNNs, including the regressive denoising DNN and the discriminative recognition DNN, into a complete multi-task structure and all the parameters can be optimized in a real joint-learning mode just from the beginning in model training. In addition, the basic multi-task structure is further explored and reorganized into a more general framework which can get substantial gains. Furthermore, noise adaptive training can also be easily incorporated within this architecture to achieve further performance improvement. Experiments on the Aurora4 task showed that the proposed approach can achieve a WER below 10% without using adaptation or sequence training, a very large and significant (more than 20% relative) improvement over a strong DNN-HMM baseline.

Although context-dependent DNN-HMM systems have achieved significant improvements over GMM-HMM systems, there still exists big performance degradation if the acoustic condition of the test data mismatches that of the training data. Hence, adaptation and adaptive training of DNN are of great research interest. Previous works mainly focus on adapting the parameters of a single DNN by regularized or selective fine-tuning, applying linear transforms to feature or hidden-layer output, or introducing vector representation of non-speech variability into the input. These methods all require relatively large number of parameters to be estimated during adaptation. In contrast, this paper employs the cluster adaptive training (CAT) framework for DNN adaptation. Here, multiple DNNs are constructed to form the bases of a canonical parametric space. During adaptation, an interpolation vector, specific to a particular acoustic condition, is used to combine the multiple DNN bases into a single adapted DNN. The DNN bases can also be constructed at layer level for more flexibility. The CAT-DNN approach was evaluated on an English switchboard task in unsupervised adaptation mode. It achieved significant WER reductions over the unadapted DNN-HMM, relative 6% to 8.5%, with only 10 parameters.

Nowadays, cloud storage systems may contain tens of thousands of servers and large scale data sets, which significantly require efficient data management scheme and query processing mechanism. To fulfill these requirements in modern data centers, the infrastructure of cloud systems, we propose FT-Index, a secondary indexing scheme for cloud system with switch-centric topology. FT-Index has a two-layer design. The upper-layer index, called global index, is distributed across different hosts in the system, while the lower-layer index, named local index, is a B+-tree for local query. We further adopt the Interval tree to reorganize the global index and propose two versions of FT-Index with different publishing methods to lower the rate of false positives and reduce the cost of forwarding queries. We provide detailed theoretical analysis on the upper bound of false positives, physical hops per query, and the relationship between them. We also conduct abundant experiments to validate the efficiency of FT-Index.