Expressions and phases for writing

这是我在读论文的过程中总结的，内容会不断地增加，慢慢完善。主要针对计算机相关专业，如果有什么错误，请留言。我是传播知识的小蜜蜂

Remarks for English expressions

lags behind, with respect to

it lags behind its non-stochastic counterparts with respect to the convergence rate.

mitigate, shortcoming

SVRG mitigates this shortcoming.

involve

Several machine learning and optimization problems involve the minimization of a smooth, convex and separable cost function.

entail

many more tasks in machine learning entail an optimization of similar form.

outgrow, pose

the volume of input data outgrows our computational capacity, posing major challenges.

incur

it incurs a prohibitive cost for very large problems.

hurdle

stochastic gradeint descent overcomes this hurdle by computing only a surrogate of the full gradient.

manyfold

the problems would be multiplied manyfold.

in the course of

the approximate ‘gradient’ of stochastic methods introduce variance in the course of the optimization.

occasionally, interleave

A full gradient computation is occasionally interleaved with the inexpensive steps of SGD.

impede

an infrequent computation of the full gradient may severely impede the progress of these variance-reduction approaches.

scheme

a variant of the popular SVRG scheme

encompass

cheapSVRG can be seen as a family of stochastic optimization schemes encompassing SVRG and vanilla SGD.

supplement

we supplement our theoretical analysis with experiments on synthetic and real data.

claim

empirical evaluation supports our claims for linear convergence.

spark

Explosive growth in data and availability of cheap computing resources have sparked increasing interest in Bing learning.

substantial

substantial recent developments

explosive

Besides, the explosive growth in volume, Big Data also has high velocity.

painful, error prone

it would be painful and error prone to write the parallel code using the low-level system promitives.

twofold

the reasons for this are twofold.

vice versa

certain

if we knew that the event is certain to happen we would receive no infomation.

intimate

there is an intimate relationship between data compression and density estimation.

aggressively

Data transmission can be reduced aggressively for energy saving without a large degradation of observation fidelity.

excessive, prohibitive

However, the optimal solution requires excessive computing and memory resources, which might be prohibitive for sensor nodes.

lightweight

Therefore, we propose a lightweight greedy algorithm to solve it.

assemble

The sink node assembles all the received samples to reproduce the evolving process of certain physical phenomenon in the monitored area.

pioneering

Several pioneering methods have been proposed to adjust the spatial sampling rata according to the statistic features of the monitored phenomenon.

inseparable

Outlier detection should be an inseparable part of any data processing routine that takes place in WSNs.

exemplify

We exemplify the benefits of our algorithm by implementing it using two different outlier detection heuristics.

accommodate

Our algorithm is flexible in that it accommodates a whole class of unsupervised outlier detection techniques.

ongoing

Data acquisition is an issue of ongoing attention for geographical information science.

be drained of

Lifetime is defined as the number of rounds until the first sensor is drained of its energy.

break down

The death of the sink node breaks down the functionality of the network.

scheme

Authors have proposed a hybrid scheme which combines the salient features of the in-network and grid-based aggregation schemes.

operational

Even if one cluster head fails, the network may still be operational.

encompass

Through experiments on 13 different testbeds, encompassing 7 platforms, 6link layers, and multiple densities and frequencies.

spur

This has spurred a substantial amount of research on wireless sensor networks over the past few years.

preserve

The second approach preserves the original information.

thoroughly

In the sequel, we thoroughly review each of the aforementioned functionalities.

be inferior to

Its performance in terms of aggregation effectiveness is largely inferior to that of the MST.

address

We do not propose any new protocols in this paper, but rather attempt to address at a higher level.

envision

Sensor networks of the future are envisioned to revolutionize the paradigm of collecting and processing information in diverse environments.

periodically

Each sensor periodically produces information as it monitors its vicinity.

replenish

Replenishing energy via replacing batteries on hundreds of nodes is infeasible.

vast

There is a vast amount of extant work on in-network aggregation in the literature.

performance gain

It can potentially lead to significant performance gains.

facilitate

Many evolving low-power sensor spatial density to enable reliable operation in the face of component node failures as well as to facilitate high spatial localization of events of interest.

mirror

Despite the existence of potential application, the conceptual importance of distributed source coding has not been mirrored in practical data compression.

mitigate

We combine scheduling with transmission power control to mitigate the effects of interference.

be fit for

TDMA is better fit for fast data collection.
Since the sink remains as the bottleneck, sending data over different paths does not reduce the schedule length.

for the purpose of

The process of monitoring structures for the purpose of damage identification is known as structural health monitoring.

proliferate, human intervention

The adoption of WSNs in advanced structural health monitoring systems has proliferated in the last few decades because of their ability to operate reliably without human intervention in inaccessible areas.

match

The task scheduler must ensure that task allocation is matched to the available energy.

superiority

We demonstrate the superiority of our linear regression algorithm.

comprise

This platform comprises three boards.

burgeon

As a burgeoning technique for signal processing, compressed sensing is being increasingly applied to wireless communications.

tackle

It has been tackled from various aspects since the outset of WSNs.

complement, be deemed as

It complements other approaches and is deemed as the most crucial mechanism to achieve energy efficient data collection for WSNs.

heavily

Although data aggregation techniques have been heavily investigated, there are still imperfections to be improved on.

complication, postpone

The complication involved in the interaction between data routing and CS-based aggregation has postponed the development on this front until very recently.

tricky

We prove the NP-completeness of MECDA in general through a tricky reduction.

validate

We report a large set of numerical results, which validate the efficacy of our heuristic.

give a brief of overview

We first give a brief overview of applying compressed sensing in networking.

coin

We coined this scheme as A in our previous work.

cardinality

We denote by n and l the cardinalities of V and E.

hurt

This is because once CS aggregation is initiated, allowing additional raw data transmission would only hurt the energy efficiency.

as a byproduct

As a byproduct, we also obtain the inapproximability of MECDA.

intractable

Other parameterized versions of CSATCP are intractable.

randomized, benchmark

A randomized algorithm is borrowed from [18] to benchmark our heuristic in large scale WSNs.

prominent

We have shown the prominent improvement in energy efficiency of our method through numerical simulations.

have capability of

WSNs are a class of ad hoc networks that has capability of self-organizing, in-network data processing, and unattended environment monitoring.

scale down

This feature scales down the cost of owning, programming and maintaining the WSNs.

prioritise

We prioritized saving sensor’s energy over latency in receiving data.

cope with

The proposed scheme can cope with abnormal sensor readings gracefully.

a wealth of

It will bring a wealth of similar benefits as distributed source coding.

merit, promising

All these desired merits make compressive sampling a promising solution to the data gathering problem in large-scale wireless sensor networks.

rescue

If some sensor nodes in a network lose a piece of crucial information, other sensor nodes may come to the rescue by providing the missing data.

categorized

Scheduling of loads has also been categorized as either job scheduling or task scheduling.

tractable

DLT provides a tractable and practical approach to load scheduling problems involved in sensing, communication and computation aspects.

bear, resemblance

Divisible load theory bears some resemblance to energy based task allocations.

baseline

A is used for baselines.

toward

Toward such a goal, we propose an energy-balanced task allocation such that the maximal energy dissipation among all sensor nodes during each period is minimized.

spark

The widespread dissemination of small-scale sensor nodes has sparked interest in a powerful new database abstraction for sensor networks.

pose a number of challenges

Query processing in large-scale sensor networks poses a number of challenges.

refine, account for

We refine our multi-query optimization algorithms to account for computational and memory limitations of sensor nodes.

be complementary to

The contribution of UMADE is complementary to this paper.

contention

Running multiple queries in such an uncooperative manner will lead to bandwidth contention and even data loss as a result of transmission collisions.

eliminate

Duplicate data requests from original queries can be eliminated as much as possible while guaranteeing the correctness of semantics of all queries.

harsh

harsh deployment environments

inject

inject them into the sensor network

preconditioning

in a preconditioning step

intimately

intimately related

forefront

the forefront of current research

premature

due to premature battery exhaustion

malfunction

sensor malfunction

depletion

a faster depletion of its battery power

residual

residual power

shift

This paradigm shifts the focus from the traditional approaches to a more data-centric approach.