Computational Science - Science topic

Thanks for sharing. Wishing you every success in your task.

Please let me know if anyone is interested to o

These topics represent some of the exciting frontiers where machine learning is making significant contributions to computational science and engineering research.

Hemorrhagic fever with renal syndrome is a non- transmissible viral zoonosis widespread in Russia. It is considered one of the most important natural focal diseases. We examined leptospirosis and tularemia as climate-dependent diseases of bacterial etiology. Leptospirosis is transmitted only non-transmissively, while tularemia, in addition to a number of non-transmissible pathways, is characterized by a facultative- transmissible pathway, and arthropods, primarily blood- sucking dipterans, act as mechanical carriers.

Among the transmissible natural focal diseases of various etiologies, we examined tick-borne encephalitis, ixodid mites-borne borreliosis (Lyme disease), mites- borne typhus (mites-borne rickettsiosis), and Crimean hemorrhagic fever.

West Nile fever is one of the most important natural focal diseases transmitted by mosquitoes. The most unfavorable situation was observed in Astrakhan, Volgograd, and Rostov regions. They account for the vast majority of cases of the disease.

A classic example of anthroponosis is malaria. The epidemiological situation at the present time can be considered favorable given that the number of imported cases is small (about a hundred per year) and local cases are rare. However, malaria is also a classic example of a disease that can quickly regain its position when control is loosened.

Model analysis of changes in the distribution of Ixodidae

mites and malaria mosquitoes showed that their habitats were influenced by the observed climate change which is significantly expanding in the northern and eastern regions, and the reduction is negligible.

The prerequisites for an increase in the incidence of these infections and their wider spread persist.

Climate-related risk factors include expansion of the areas of arthropod vectors and an increase in their numbers, as well as a similar increase in the number and expansion of the areas of vertebrates, mainly murine rodents, which are reservoirs of natural focal infections and carriers of vectors in nature. A new risk factor is the importation of exotic vectors into the territory of Russia, as well as their rooting, the importance of which increases with the expected warming.

Countermeasures (adaptation measures) against climate-dependent infections include prevention (vaccination), strengthening monitoring of the species composition and number of vectors and reservoirs of infections, and increasing the scale and effectiveness of combating them. These measures also include personal protective equipment against them. The enhancement of the effectiveness of these measures should be based not only on the actual improvement of these funds but also on the improvement of the sanitary and epidemiological

We have recently published two papers and proposed different adaptive methods for the hyperparameters tuning as follows

1.

I will get back to you soon.

Dear Mirlan,

My congratulations on your graduation. Regarding your question about future studies at ETH, I have looked at the outline of the courses mentioned in your question.

You are probably familiar with these sites but I have included the links just for the documentation:

Advanced numerical methods

Optimization for data science

Computational statistics

Advanced systems lab

Based on the analysis of the above, I will probably choose the Advanced Numerical Methods and the Advanced Systems Lab. (I really like these courses and I think they are very useful regardless of the future specialization).

I wonder how the current situation (related to Covid-19) with online courses is evolving at ETH? Will this constraint change schedules and plans?

In any case, my best wishes for the success of your program.

Kind Regards

This was only a letter addressed to ResearchGate answering to its query, but by no means to be open to the public. So please, drop it.

Thanks Dr. Seiti. But, i could not get any of the two papers you have referred as the full texts are not available in RG.

I have put a request to the authors and hope to get copies of the papers soon to verify your suggestions.

regards,

B. K. Tripathy

Two route to answer here.

The first is general green technology in general. Sustainability initiatives require reducing the energy intensity of industrial and consumer applications. Computing is part of that. Greener computing reduces the energy intensity of economic activities improving the energy intensity of economies (i.e. increasing the GDP per unit of energy used.) This has the same general impact as environmental sustainability issues in general.

Second is in terms of operations cost. For large data center or supercomputers the energy intensity is a big concern. Let's take round order of magnitude numbers. A top supercomputer today costs on the order of 100M$. (or maybe a little less, I overestimate a bit to make THIS calculation conservative.) The power required to run the machine is on the order to 10 MW. If the machine has a useful life of say 3 years. This is 26,000 hours. Now if you are lucky you can buy energy for $0.1/kWhr. That means your machine costs $1,000 per hour to run just for the electricity. Over the life of the machine you will spend 26M$ just for electricity! This has two BIG effects. First the total cost of ownership of supercomputers (or datacenters) is greatly affected by the energy use. Second, because of the high cost of energy, the lifetime of machines are shorter. Machines still work after 3 years but the cost to operate them are too high in comparison to value when the newer machines are available in three years and the old ones are too expensive to operate. Let me restate this in bold: supercomputers are turned off not because they end their useful life but because they use too much energy in relation to other options.

Green computing means less energy per computation which allows you to buy more computer per budget dollar on a lifetime cost basis and allows you to economically run the machine longer.

In fact it is so important that the logic runs in reverse. For example, early scoping studies for the US exascale effort REQUIRED the power of an exaflop machine to be less than 20 MW. Straight scaling from the petaflop era would have given a machine with a power well above 100 MW. So from the start, the exaflop project REQUIRED planners to plan for an increase in computing efficiency of an order of magnitude. To state it another way, green computing is a prerequisite to play the supercomputing game today. You must be this green to rider this ride.

Note: this discussion is not quite the same as the discussion of the "power wall" and the failing of Dennard scaling in HPC. There the issue is the inability to dissipate heat on a chip. However, the two problems scale similarly and the solution to one helps the solution to the other.

1) They do not provide in their API the same results as in Google Live.

2) The results are just an estimate of the total. If you navigate through the results pages you will see that for some queries there are actually less results than the number they show you.

3) Another point is that it is very difficult to compute NGD, you can check this thesis (chapter 4 (also 3)) as an alternative (method CBM): http://www.dbd.puc-rio.br/pergamum/tesesabertas/1012681_2014_completo.pdf

You might use these ones;

There are many examples from beginning to advance!

Your question is kind of vague...

If you mean Quantum chemical/Computational chemistry studies of carbon in its various allotropic forms, the answer is yes. Otherwise, you need to clarify your question.

For the QM part you may start here:

Basiuk VA, Rybak-Akimova EV, & Basiuk EV (2017) Graphene oxide and nanodiamond: same carboxylic groups, different complexation properties. RSC Adv. 7:17442-17450.

Dear Muhammad Gulraj,

As Dr. Ramon López de Mántaras pointed out Bee or Ant colonies foraging is a behavior that is simulated to solve optimization problems. Grey Wolf Optimizer(GWO) algorithm mimics the leadership hierarchy and hunting mechanism of grey wolves in nature.  Particle Swarm Optimization (PSO) simulates the social behavior of bird flocking or fish schooling which is used to solve different optimization problems. For more information please see the following links:

Regards,

Yes.

Reading in general is good, but an important part of original research is, based on knowledge of a field, come up yourself or with some little help with research topics to investigate. An important part of a Ph.D. research is finding the important, relevant, challenging subjects you can develop on, solve, advance.

If you get direct advice from others you can be a technician of some sort, but not a complete researcher!

I think there is no such module available in python. We need to write..

I think if you use the origin packaging from grlib, it works perfectly. Although there are x propagation in early simulation, it is ok. If the designated program runs on it, it can give the desirable results.

That is my experience using leon3.

Cheers

I require datasets number of instances(ROWS) more then 10K. Further positive values means positive values of attributes.

Thanks

Non, it's parallel licence 

Spontaneous emission (SE) rates from semiconductors is interestingly approximated by a dipole radiation using classical electrodynamics even though SE actually requires quantum electrodynamics explanation (This was shown as early as in 1999 by Xu et al). That is why I found many papers which use a dipole excitation. But the question is about validity of monopole sources and single pulse/continuous wave at femto second timescales.

Dear Laouid,

You will find excellent information at the following links

With my best regards

Prof. Bachir ACHOUR

Dear Prof.  Andriy O. Borisyuk(Borysyuk) !

Many thanks for your suggestion !

I don't fully understand your question, but:

1) the PDF file is from 2009 ! This was an ancient times of GPGPU ! Maybe this is more up-to-date https://wiki.postgresql.org/wiki/PGStrom

2) you should decide on which level you will write your programs: at the CUDA level or at SQL level?

The work by Osorio Saraz et al. (2012) about "Fuzzy modeling applied to the welfare of poultry farms workers" may be suitable to start with the idea.

Jairo Alexander Osorio Saraz; Leonardo Schiassi; Tadayuki Yanagi Junior; Flávio Alves Damasceno; Neiton Silva Machado, (2012). Fuzzy modeling applied to the welfare of poultry farms workers, DYNA. Revista de la Facultad de Minas, Vol 79, No 174; pp. 127-135.

In addition, consider searching from

both sides. When two opposing vertices

from different starting points meet,

you can stop. For example, you can

start from one side with positive

counting 1,2,3,4... from the other

side with -1,-2,-3,-4... When an edge

appears with a positive and negative

node on either side, you can construct

the path by stepping back in descending

and ascending order.

You can also use 2 marks instead

of a single positive/negative.

Regards,

Joachim

You question is one of the most used in this time.

You see it from holistic commercial side. I mean you can start hope in some cases (called Quantum Circuits). They are quite realistic in special cases.

But in general I suppose the Quantum Computer can't be realized in future "as one unified concept".

As actual state is we need special mathematical algorithms (HADAMARD, DEUTSCH, SHOR) to get parallel functioning and a reverse fixing of states of a Quantum Circuit to get that fascinating "parallel computing". In reality many such special physical circuits are found and all claim to be the first "quantum computer".

So my actual state of knowledge is that a unified quantum computer cant be, because we can't rebuilt the physical world of quants and steer them as we need them for our results. We have to create some very special circuits and surround them by our mathematically defined algorithms to get some effects like our particles have in Quantum Mechanics.

Sure is too that not in all cases a Quantum Computer (Quantum Circuit) is better than our classical computer. That special positive advantages are f. i. : factoring of numbers and combined "cryptomysing" of data.    

So there will be a lot to do for algorithmic and mathematics to use quantum circuits for more cases in commercial purposes.     

ok, good.  Then see Arxiv 1504.01964 for a procedure, which calculates y directly, without going through the intermediate calculation of exp(x).  It is described in the Arxiv note as being the solution of y = log(W(exp(x))) but it is mathematically equivalent to y = x - W(exp(x)).  And the algorithm does not require calculation of exp(x) intermediate.  Please let me know if you have trouble implementing in Matlab.

Maybe it is even sufficient to have a solver that only supports OR but not AND. The only prerequisite is to also have negation. Then you can use De Morgan's Law to remove all ANDs. Together with Frédéric's answer this might help to use existing solvers.

STDP idea deserves attention. However, additive and multiplicative models changes synapse weights are not perfect. They do not take into account the system-wide factors specific to neural networks. Now to the question about the typical of distribution of synapse weights. You should keep in mind that the typical of distribution weights of synapses carry some genetic information. Without knowledge of the specific tasks assigned to the neural network, the typical distribution of synapse weights can not be justified.

You might be interested in this paper (we used T-invariants as a memory optimization heuristics during state exploration):

R. Carvajal-Schiaffino, G. Delzanno and G. Chiola.

Combining Structural and Enumerative Techniques for the Validation of Bounded Petri Nets.

In T. Margaria and W. Yi, editors. Tools and Algorithms for the Construction and Analysis of Systems. TACAS 2001.

Lecture Notes in Computer Science 2031, Springer. 2001.

In general it is not possible to decide

whether two algorithms do the same,

see Rice's theorem.

Regards,

Joachim

S is a searchable encryption of a keyword, i.e. the output of PEKS. The encryption has two parts, the first part (g^r) is referred as A, the second part (H2(t)) is referred as B in TEST.

I am not sure if it is exactly what you need, but you can try MolGen (http://molgen.de/?src=documents/molgenonline).

There is no theoretical reason. The most compelling practical reason for this is cost. Maybe a minor practical reason is that is hard to implement a fast search for a larger cache.

The thing is that fast memory is very expensive. Otherwise we would use computers with the number of registers equivalent to several gigabytes of data. Because the processor logic would be quite complex for billions of registers (and thus expensive) we are using a L1 cache. This cache is still quite fast, but having a larger size of this kind of fast memory would again cost a lot. Therefore, the trade-off is again a small size. In the beginning of computing it would only take about 2 CPU cycles to access RAM (which by today's standards is also really small). With every increase in CPU speed another level of cache is needed to compensate for speed differences between CPU registers and RAM. Large and slow memory costs as much as fast and small memory. This is always a trade-off between size, speed, and money. You cannot optimize for all three at the same time. The main reason to have a cache is to hide latency to the RAM (or to the L3 cache, or to the L2 cache). Sure, this only works for programs that adhere to memory locality. Otherwise, you would have roughly 200 CPU cycles without any work; just waiting for data from the RAM.

I am sorry, I should have made it a bit clear.

Say, I am trying to identify the communities in an unsupervised manner and for that  I am trying to maximize the modularity. Now, I get different number of communities with different nodes even at a single resolution parameter. The question arises which of the communities is the best i.e. is there any statistical criteria which can lead me to find that number? 

Moreover, the choice of resolution parameter itself is a question mark.

No, it will not lead to errors in the results, but first you must be sure to use a computer with enough memory to handle such "huge" datasets. The fact of having several variables with the same name will only use more memory. I suggest you to try another approach such as parallelizing your calculation, storing your datasets in external tables and workspaces and loading them only when they are necessary, optimizing the code.             

Hello, Cihan! I suggest you to take a look at these papers:

1) http://epubs.siam.org/doi/abs/10.1137/0208040

2) http://www.sciencedirect.com/science/article/pii/S0747717110001136

3) http://arxiv.org/abs/1401.1773

4) http://www.tandfonline.com/doi/abs/10.1080/03081087.2012.743540#.U9_aO2PEiMo

Best wishes and good luck with your research!

I am not clear about the purpose of this rotation. But if you are interested in the potential energy surface for this torsional angle rotation you can do a relaxed PES scan.

Have you tried making a tight example? That is, an infinite set of instances that hit your approximation factor yet?

I am working on the makespan problem on unrelated parallel machines these days, so I can suggest some ideas as your problem is very related to this (not the same though). If experimental and not formal results are wanted, one thing I can suggest is just generating permutations of a schedule you know will have an optimal load you know (say m). A big thing to deal with in a scheduling problem is what order jobs are processed. For example, the classic list scheduling algorithm by Graham does some pretty nasty things if you force it to pick bad choices then try to squeeze a job on a machine that is already balanced. That's how you get the tight example.

One thing I can suggest is looking at your algorithm, and the properties it has, and prove it has a certain approximation factor based on the characteristics of your algorithm. Usually knowing the approximation factor typically requires finding the flaws in the algorithm by finding a provably good bound you know about the scheduling problem.

But yes, for experimental results, try permuting a set of jobs where about half are even, half are odd. Usually even and odd job lengths (with difference greater than 2) can do funky things to a schedule sometimes. I can't suggest any standard instances beyond what I said, as I usually go about this formally, since an approximation factor is something somebody can prove usually.

Hope this helps!

It depends on what level of independence you expect from the two programs. Generally the first answer would be "yes", since the usual level of independence is good for most cases. I.e. if just run on two cores, one program should not crash the other (provided the program and the OS are free of bugs). The ways in which these programms may interact is only relevant if you need very clear guarantees, e.g. real-time behavior or safety-related functions.

In case hard real-time behavior is required for example the question cannot be answered without knowing the exact hardware. You mention that the programs do not use shared resources, but this is close to impossible to realize in realistic hardware. Most times the cores of a multiprocessor share caches or the connection to main memory. So in case one programm accesses the main memory, the other may have to wait for the access to wait for the first program to finish, which can change the execution time. Or the programs may share cache lines which can also influence performance. It would even be possible that one program drives up the heat, so that the OS has to activate a power saving mode for both cores which then again degrades performance for the other program.

The point is, that the concept of two programs sharing no resources is an illusion provided by the OS and the hardware. Some resources will always be shared (and even if it is only because of physical proximity of the cores, making them share a common heat-sink). In most cases however this illusion is performed so well, that one can actually assume independence of the programs. And the cases where one has to look beyond that illusion are rare and require very special methods of development.

As somebody who has actually taught a university course in Analysis of Algorithms, I felt it may be good for me to give a "stab" at answering this one.

To begin, we must understand that programs and algorithms are not the same thing. One is an implementation, the other is a mathematical description. One is associated with a machine, the other with a computational model. Why do we care about analysis of algorithms? This question is trivial to answer. We care about having efficient algorithms in terms of time and space. The computational models at the present time reflect quite well (and very rarely not) the behaviour of an algorithm when it is implemented. Remember, the interest is in algorithms typically. We prove theorems and analyze algorithms, not their implementations.

So why would we care about this in software tools? As I said, computational models seem to be pretty accurate to how they perform (and often is of research interest to some). For example, the RAM model is extremely good for most cases in analyzing an algorithm in the worst-case. Typically we are interested in the behaviour of the algorithm in specific cases. Why would we care about this? Because, if we know the algorithm will have a certain asymptotic behaviour, then we know in the worst-case it cannot do any worse than that. The other factor to consider is asymptotics. If we know the behaviour as the input gets large, we have a better idea of how it will compare in complexity with other algorithms.

Not just in bioinformatics (which relies a lot on algorithms), but any area of Computing needs to consider algorithms as formal descriptions and ensuring they are correct and terminate with a certain time-complexity. These are not just "abstract ideas" that serve no purpose. They lay the ground work of almost everything we do in Theoretical Computer Science (core of CS) when it comes to Algorithms. Once you have time-complexities, we can compare algorithms, and even categorize them and ask questions about them. Many bioinformatics problems in particular can be NP-hard, so this will be even more important when considering an application.

Hope this helps!

@James

I didn't know about I_MPI_WAIT_MODE yet. But, from the manual I take it that it should be set to 'on'. 'off' is the default value and will always do active polling. There is one restriction, though: It does not support InfiniBand devices.

I have been looking into a similar option as you are describing, but using mostly portable FORTRAN for sleeping: We are using the Intel compiler and also the software has to run with Windows, not just Linux. Here is what I do:

do

call SLEEPQQ(60000) ! sleep one minute

call MPI_IPROBE(MPI_ANY_SOURCE, MPI_ANY_TAG, &

MPI_COMM_WORLD, flag, tmp, ierror)

if (flag .eq. .true.) exit

end do

call MPI_Finalize()

Then the master process just needs to send any message and the sleeping thread will eventually finish. This works. But, I am curious if there is a more elegant solution to this problem that does not involve active polling.

@Cedrick

MPI can only use shared memory for communication which will make communication faster. But, because of the programming model each process will have to load the same data. There is no shared memory for data in MPI. This is the advantage of OpenMP which will decrease memory footprint. And this is the only reason why we consider a hybrid approach.

We will always keep the pure MPI version and pure OpenMP version of our software as well.

BTW, I know about processor pinning. But, this is not the problem I am concerned with. It is quite easy to figure out for Intel MPI and Intel OpenMP. The real question is how do I make my application smart enough to configure itself for hybrid use. I thought about this a lot and I think that it is definitely necessary to first start one MPI process per core. Then I need to figure out how many MPI processes should participate in the computation and which ones those are. These MPI processes will later split work on OpenMP threads. The only question is what to do with the remaining MPI processes that are not necessary for computation.

did you try checking for algorithms for microstructure creation in 2D (e.g. based on Voronoi tessellation/Delaunay triangulation)? In that case you get a distribution of radii (it is unavoidable.. otherwise you have to put in each point a particle with a radius equal to the mid distance between the two closest points on the mesh.

Probably you can find some useful information checking at Surface Evolver as well (http://www.susqu.edu/brakke/evolver/evolver.html).

The name is Verlet list (cell list). A quick intro and some references to start are here: http://en.wikipedia.org/wiki/Cell_lists.

You can write whole papers using SWeave, which integrates literate programming with the R language and statistics framework http://users.stat.umn.edu/~geyer/Sweave/ There are several examples of integration of R and LaTeX down the page (My Examples)