Search Results

Search found 8262 results on 331 pages for 'optimization algorithm'.

Page 185/331 | < Previous Page | 181 182 183 184 185 186 187 188 189 190 191 192  | Next Page >

  • How do I do high quality scaling of a image?

    - by pbhogan
    I'm writing some code to scale a 32 bit RGBA image in C/C++. I have written a few attempts that have been somewhat successful, but they're slow and most importantly the quality of the sized image is not acceptable. I compared the same image scaled by OpenGL (i.e. my video card) and my routine and it's miles apart in quality. I've Google Code Searched, scoured source trees of anything I thought would shed some light (SDL, Allegro, wxWidgets, CxImage, GD, ImageMagick, etc.) but usually their code is either convoluted and scattered all over the place or riddled with assembler and little or no comments. I've also read multiple articles on Wikipedia and elsewhere, and I'm just not finding a clear explanation of what I need. I understand the basic concepts of interpolation and sampling, but I'm struggling to get the algorithm right. I do NOT want to rely on an external library for one routine and have to convert to their image format and back. Besides, I'd like to know how to do it myself anyway. :) I have seen a similar question asked on stack overflow before, but it wasn't really answered in this way, but I'm hoping there's someone out there who can help nudge me in the right direction. Maybe point me to some articles or pseudo code... anything to help me learn and do. Here's what I'm looking for: 1. No assembler (I'm writing very portable code for multiple processor types). 2. No dependencies on external libraries. 3. I am primarily concerned with scaling DOWN, but will also need to write a scale up routine later. 4. Quality of the result and clarity of the algorithm is most important (I can optimize it later). My routine essentially takes the following form: DrawScaled( uint32 *src, uint32 *dst, src_x, src_y, src_w, src_h, dst_x, dst_y, dst_w, dst_h ); Thanks! UPDATE: To clarify, I need something more advanced than a box resample for downscaling which blurs the image too much. I suspect what I want is some kind of bicubic (or other) filter that is somewhat the reverse to a bicubic upscaling algorithm (i.e. each destination pixel is computed from all contributing source pixels combined with a weighting algorithm that keeps things sharp. EXAMPLE: Here's an example of what I'm getting from the wxWidgets BoxResample algorithm vs. what I want on a 256x256 bitmap scaled to 55x55. And finally: the original 256x256 image

    Read the article

  • To sample or not to sample...

    - by [email protected]
    Ideally, we would know the exact answer to every question. How many people support presidential candidate A vs. B? How many people suffer from H1N1 in a given state? Does this batch of manufactured widgets have any defective parts? Knowing exact answers is expensive in terms of time and money and, in most cases, is impractical if not impossible. Consider asking every person in a region for their candidate preference, testing every person with flu symptoms for H1N1 (assuming every person reported when they had flu symptoms), or destructively testing widgets to determine if they are "good" (leaving no product to sell). Knowing exact answers, fortunately, isn't necessary or even useful in many situations. Understanding the direction of a trend or statistically significant results may be sufficient to answer the underlying question: who is likely to win the election, have we likely reached a critical threshold for flu, or is this batch of widgets good enough to ship? Statistics help us to answer these questions with a certain degree of confidence. This focuses on how we collect data. In data mining, we focus on the use of data, that is data that has already been collected. In some cases, we may have all the data (all purchases made by all customers), in others the data may have been collected using sampling (voters, their demographics and candidate choice). Building data mining models on all of your data can be expensive in terms of time and hardware resources. Consider a company with 40 million customers. Do we need to mine all 40 million customers to get useful data mining models? The quality of models built on all data may be no better than models built on a relatively small sample. Determining how much is a reasonable amount of data involves experimentation. When starting the model building process on large datasets, it is often more efficient to begin with a small sample, perhaps 1000 - 10,000 cases (records) depending on the algorithm, source data, and hardware. This allows you to see quickly what issues might arise with choice of algorithm, algorithm settings, data quality, and need for further data preparation. Instead of waiting for a model on a large dataset to build only to find that the results don't meet expectations, once you are satisfied with the results on the initial sample, you can  take a larger sample to see if model quality improves, and to get a sense of how the algorithm scales to the particular dataset. If model accuracy or quality continues to improve, consider increasing the sample size. Sampling in data mining is also used to produce a held-aside or test dataset for assessing classification and regression model accuracy. Here, we reserve some of the build data (data that includes known target values) to be used for an honest estimate of model error using data the model has not seen before. This sampling transformation is often called a split because the build data is split into two randomly selected sets, often with 60% of the records being used for model building and 40% for testing. Sampling must be performed with care, as it can adversely affect model quality and usability. Even a truly random sample doesn't guarantee that all values are represented in a given attribute. This is particularly troublesome when the attribute with omitted values is the target. A predictive model that has not seen any examples for a particular target value can never predict that target value! For other attributes, values may consist of a single value (a constant attribute) or all unique values (an identifier attribute), each of which may be excluded during mining. Values from categorical predictor attributes that didn't appear in the training data are not used when testing or scoring datasets. In subsequent posts, we'll talk about three sampling techniques using Oracle Database: simple random sampling without replacement, stratified sampling, and simple random sampling with replacement.

    Read the article

  • ASP.NET MVC 4: Short syntax for script and style bundling

    - by DigiMortal
    ASP.NET MVC 4 introduces new methods for style and scripts bundling. I found something brilliant there I want to introduce you. In this posting I will show you how easy it is to include whole folder with stylesheets or JavaScripts to your page. I’m using ASP.NET MVC 4 Internet Site template for this example. When we open layout pages located in shared views folder we can see something like this in layout file header: <link href="@System.Web.Optimization.BundleTable.Bundles.ResolveBundleUrl("~/Content/css")" rel="stylesheet" type="text/css" />    <link href="@System.Web.Optimization.BundleTable.Bundles.ResolveBundleUrl("~/Content/themes/base/css")" rel="stylesheet" type="text/css" />    <script src="@System.Web.Optimization.BundleTable.Bundles.ResolveBundleUrl("~/Scripts/js")"></script> Let’s take the last line and modify it so it looks like this: <script src="/Scripts/js"></script> After saving the layout page let’s run browser and see what is coming in over network. As you can see the request to folder ended up with result code 200 which means that request was successful. 327.2KB was received and it is not mark-up size for error page or directory index. Here is the body of response: I scrolled down to point where one script ends and another one starts when I made the screenshot above. All scripts delivered with ASP.NET MVC project templates start with this green note. So now we can be sure that the request to scripts folder ended up with bundled script and not with something else. Conclusion Script and styles bundling uses currently by default long syntax where bundling is done through Bundling class. We can still avoid those long lines and use extremely short syntax for script and styles bundling – we just write usual script or link tag and give folder URL as source. ASP.NET MVC 4 is smart enough to combine styles or scripts when request like this comes in.

    Read the article

  • How to tell if SPARC T4 crypto is being used?

    - by danx
    A question that often comes up when running applications on SPARC T4 systems is "How can I tell if hardware crypto accleration is being used?" To review, the SPARC T4 processor includes a crypto unit that supports several crypto instructions. For hardware crypto these include 11 AES instructions, 4 xmul* instructions (for AES GCM carryless multiply), mont for Montgomery multiply (optimizes RSA and DSA), and 5 des_* instructions (for DES3). For hardware hash algorithm optimization, the T4 has the md5, sha1, sha256, and sha512 instructions (the last two are used for SHA-224 an SHA-384). First off, it's easy to tell if the processor T4 crypto instructions—use the isainfo -v command and look for "sparcv9" and "aes" (and other hash and crypto algorithms) in the output: $ isainfo -v 64-bit sparcv9 applications crc32c cbcond pause mont mpmul sha512 sha256 sha1 md5 camellia kasumi des aes ima hpc vis3 fmaf asi_blk_init vis2 vis popc These instructions are not-privileged, so are available for direct use in user-level applications and libraries (such as OpenSSL). Here is the "openssl speed -evp" command shown with the built-in t4 engine and with the pkcs11 engine. Both run the T4 AES instructions, but the t4 engine is faster than the pkcs11 engine because it has less overhead (especially for smaller packet sizes): t-4 $ /usr/bin/openssl version OpenSSL 1.0.0j 10 May 2012 t-4 $ /usr/bin/openssl engine (t4) SPARC T4 engine support (dynamic) Dynamic engine loading support (pkcs11) PKCS #11 engine support t-4 $ /usr/bin/openssl speed -evp aes-128-cbc # t4 engine used by default . . . The 'numbers' are in 1000s of bytes per second processed. type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes aes-128-cbc 487777.10k 816822.21k 986012.59k 1017029.97k 1053543.08k t-4 $ /usr/bin/openssl speed -engine pkcs11 -evp aes-128-cbc engine "pkcs11" set. . . . The 'numbers' are in 1000s of bytes per second processed. type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes aes-128-cbc 31703.58k 116636.39k 350672.81k 696170.50k 993599.49k Note: The "-evp" flag indicates use the OpenSSL "EnVeloPe" API, which gives more accurate results. That's because it tells OpenSSL to use the same API that external programs use when calling OpenSSL libcrypto functions, evp(3openssl). DTrace Shows if T4 Crypto Functions Are Used OK, good enough, the isainfo(1) command shows the instructions are present, but how does one know if they are being used? Chi-Chang Lin, who works on Oracle Solaris performance, wrote a Dtrace script to show if T4 instructions are being executed. To show the T4 instructions are being used, run the following Dtrace script. Look for functions named "t4" and "yf" in the output. The OpenSSL T4 engine uses functions named "t4" and the PKCS#11 engine uses functions named "yf". To demonstrate, I'll first run "openssl speed" with the built-in t4 engine then with the pkcs11 engine. The performance numbers are not valid due to dtrace probes slowing things down. t-4 # dtrace -Z -n ' pid$target::*yf*:entry,pid$target::*t4_*:entry{ @[probemod, probefunc] = count();}' \ -c "/usr/bin/openssl speed -evp aes-128-cbc" dtrace: description 'pid$target::*yf*:entry' matched 101 probes . . . dtrace: pid 2029 has exited libcrypto.so.1.0.0 ENGINE_load_t4 1 libcrypto.so.1.0.0 t4_DH 1 libcrypto.so.1.0.0 t4_DSA 1 libcrypto.so.1.0.0 t4_RSA 1 libcrypto.so.1.0.0 t4_destroy 1 libcrypto.so.1.0.0 t4_free_aes_ctr_NIDs 1 libcrypto.so.1.0.0 t4_init 1 libcrypto.so.1.0.0 t4_add_NID 3 libcrypto.so.1.0.0 t4_aes_expand128 5 libcrypto.so.1.0.0 t4_cipher_init_aes 5 libcrypto.so.1.0.0 t4_get_all_ciphers 6 libcrypto.so.1.0.0 t4_get_all_digests 59 libcrypto.so.1.0.0 t4_digest_final_sha1 65 libcrypto.so.1.0.0 t4_digest_init_sha1 65 libcrypto.so.1.0.0 t4_sha1_multiblock 126 libcrypto.so.1.0.0 t4_digest_update_sha1 261 libcrypto.so.1.0.0 t4_aes128_cbc_encrypt 1432979 libcrypto.so.1.0.0 t4_aes128_load_keys_for_encrypt 1432979 libcrypto.so.1.0.0 t4_cipher_do_aes_128_cbc 1432979 t-4 # dtrace -Z -n 'pid$target::*yf*:entry{ @[probemod, probefunc] = count();}   pid$target::*yf*:entry,pid$target::*t4_*:entry{ @[probemod, probefunc] = count();}' \ -c "/usr/bin/openssl speed -engine pkcs11 -evp aes-128-cbc" dtrace: description 'pid$target::*yf*:entry' matched 101 probes engine "pkcs11" set. . . . dtrace: pid 2033 has exited libcrypto.so.1.0.0 ENGINE_load_t4 1 libcrypto.so.1.0.0 t4_DH 1 libcrypto.so.1.0.0 t4_DSA 1 libcrypto.so.1.0.0 t4_RSA 1 libcrypto.so.1.0.0 t4_destroy 1 libcrypto.so.1.0.0 t4_free_aes_ctr_NIDs 1 libcrypto.so.1.0.0 t4_get_all_ciphers 1 libcrypto.so.1.0.0 t4_get_all_digests 1 libsoftcrypto.so.1 rijndael_key_setup_enc_yf 1 libsoftcrypto.so.1 yf_aes_expand128 1 libcrypto.so.1.0.0 t4_add_NID 3 libsoftcrypto.so.1 yf_aes128_cbc_encrypt 1542330 libsoftcrypto.so.1 yf_aes128_load_keys_for_encrypt 1542330 So, as shown above the OpenSSL built-in t4 engine executes t4_* functions (which are hand-coded assembly executing the T4 AES instructions) and the OpenSSL pkcs11 engine executes *yf* functions. Programmatic Use of OpenSSL T4 engine The OpenSSL t4 engine is used automatically with the /usr/bin/openssl command line. Chi-Chang Lin also points out that if you're calling the OpenSSL API (libcrypto.so) from a program, you must call ENGINE_load_built_engines(), otherwise the built-in t4 engine will not be loaded. You do not call ENGINE_set_default(). That's because "openssl speed -evp" test calls ENGINE_load_built_engines() even though the "-engine" option wasn't specified. OpenSSL T4 engine Availability The OpenSSL t4 engine is available with Solaris 11 and 11.1. For Solaris 10 08/11 (U10), you need to use the OpenSSL pkcs311 engine. The OpenSSL t4 engine is distributed only with the version of OpenSSL distributed with Solaris (and not third-party or self-compiled versions of OpenSSL). The OpenSSL engine implements the AES cipher for Solaris 11, released 11/2011. For Solaris 11.1, released 11/2012, the OpenSSL engine adds optimization for the MD5, SHA-1, and SHA-2 hash algorithms, and DES-3. Although the T4 processor has Camillia and Kasumi block cipher instructions, these are not implemented in the OpenSSL T4 engine. The following charts may help view availability of optimizations. The first chart shows what's available with Solaris CLIs and APIs, the second chart shows what's available in Solaris OpenSSL. Native Solaris Optimization for SPARC T4 This table is shows Solaris native CLI and API support. As such, they are all available with the OpenSSL pkcs11 engine. CLIs: "openssl -engine pkcs11", encrypt(1), decrypt(1), mac(1), digest(1), MD5sum(1), SHA1sum(1), SHA224sum(1), SHA256sum(1), SHA384sum(1), SHA512sum(1) APIs: PKCS#11 library libpkcs11(3LIB) (incluDES Openssl pkcs11 engine), libMD(3LIB), and Solaris kernel modules AlgorithmSolaris 1008/11 (U10)Solaris 11Solaris 11.1 AES-ECB, AES-CBC, AES-CTR, AES-CBC AES-CFB128 XXX DES3-ECB, DES3-CBC, DES2-ECB, DES2-CBC, DES-ECB, DES-CBC XXX bignum Montgomery multiply (RSA, DSA) XXX MD5, SHA-1, SHA-256, SHA-384, SHA-512 XXX SHA-224 X ARCFOUR (RC4) X Solaris OpenSSL T4 Engine Optimization This table is for the Solaris OpenSSL built-in t4 engine. Algorithms listed above are also available through the OpenSSL pkcs11 engine. CLI: openssl(1openssl) APIs: openssl(5), engine(3openssl), evp(3openssl), libcrypto crypto(3openssl) AlgorithmSolaris 11Solaris 11SRU2Solaris 11.1 AES-ECB, AES-CBC, AES-CTR, AES-CBC AES-CFB128 XXX DES3-ECB, DES3-CBC, DES-ECB, DES-CBC X bignum Montgomery multiply (RSA, DSA) X MD5, SHA-1, SHA-256, SHA-384, SHA-512 XX SHA-224 X Source Code Availability Solaris Most of the T4 assembly code that called the new T4 crypto instructions was written by Ferenc Rákóczi of the Solaris Security group, with assistance from others. You can download the Solaris source for this and other parts of Solaris as a few zip files at the Oracle Download website. The relevant source files are generally under directories usr/src/common/crypto/{aes,arcfour,des,md5,modes,sha1,sha2}}/sun4v/. and usr/src/common/bignum/sun4v/. Solaris 11 binary is available from the Oracle Solaris 11 download website. OpenSSL t4 engine The source for the OpenSSL t4 engine, which is based on the Solaris source above, is viewable through the OpenGrok source code browser in directory src/components/openssl/openssl-1.0.0/engines/t4 . You can download the source from the same website or through Mercurial source code management, hg(1). Conclusion Oracle Solaris with SPARC T4 provides a rich set of accelerated cryptographic and hash algorithms. Using the latest update, Solaris 11.1, provides the best set of optimized algorithms, but alternatives are often available, sometimes slightly slower, for releases back to Solaris 10 08/11 (U10). Reference See also these earlier blogs. SPARC T4 OpenSSL Engine by myself, Dan Anderson (2011), discusses the Openssl T4 engine and reviews the SPARC T4 processor for the Solaris 11 release. Exciting Crypto Advances with the T4 processor and Oracle Solaris 11 by Valerie Fenwick (2011) discusses crypto algorithms that were optimized for the T4 processor with the Solaris 11 FCS (11/11) and Solaris 10 08/11 (U10) release. T4 Crypto Cheat Sheet by Stefan Hinker (2012) discusses how to make T4 crypto optimization available to various consumers (such as SSH, Java, OpenSSL, Apache, etc.) High Performance Security For Oracle Database and Fusion Middleware Applications using SPARC T4 (PDF, 2012) discusses SPARC T4 and its usage to optimize application security. Configuring Oracle iPlanet WebServer / Oracle Traffic Director to use crypto accelerators on T4-1 servers by Meena Vyas (2012)

    Read the article

  • Awesome new feature for HCC

    - by Steve Tunstall
    I've talked about HCC (Hybrid Columnar Compression) before. This is Oracle's built-in compression feature, free of charge in 11Gr2, that allows a CRAZY amount of compression on historical data inside an Oracle database. It only works if the database is being stored in a ZFSSA, Exadata or Axiom. You can read all about it in this whitepaper, which shows the huge value of HCC when used with the ZFSSA. http://www.oracle.com/technetwork/articles/servers-storage-admin/perf-hybrid-columnar-compression-1689701.html Now, even better, Oracle has announced  a great new feature in Oracle 12c called "Automatic Data Optimization". This allows one to setup HCC to AUTOMATICALLY compress data AS IT AGES.  So this is now ILM all built into the Oracle database. It's free for crying out loud. It just needs to be sitting on Oracle storage, such as the ZFSSA, Exadata or Axiom.  Read about ADO here: http://www.oracle.com/technetwork/database/automatic-data-optimization-wp-12c-1896120.pdf?ssSourceSiteId=ocomen

    Read the article

  • When there's no TCO, when to worry about blowing the stack?

    - by Cedric Martin
    Every single time there's a discussion about a new programming language targetting the JVM, there are inevitably people saying things like: "The JVM doesn't support tail-call optimization, so I predict lots of exploding stacks" There are thousands of variations on that theme. Now I know that some language, like Clojure for example, have a special recur construct that you can use. What I don't understand is: how serious is the lack of tail-call optimization? When should I worry about it? My main source of confusion probably comes from the fact that Java is one of the most succesful languages ever and quite a few of the JVM languages seems to be doing fairly well. How is that possible if the lack of TCO is really of any concern?

    Read the article

  • How can calculus and linear algebra be useful to a system programmer?

    - by Victor
    I found a website saying that calculus and linear algebra are necessary for System Programming. System Programming, as far as I know, is about osdev, drivers, utilities and so on. I just can't figure out how calculus and linear algebra can be helpful on that. I know that calculus has several applications in science, but in this particular field of programming I just can't imagine how calculus can be so important. The information was on this site: http://www.wikihow.com/Become-a-Programmer Edit: Some answers here are explaining about algorithm complexity and optimization. When I made this question I was trying to be more specific about the area of System's Programming. Algorithm complexity and optimization can be applied to any area of programming not just System's Programming. That may be why I wasn't able to came up with such thinking at the time of the question.

    Read the article

  • SH404SEF URLs in Joomla 1.5

    - by Tao Bellamine
    I have two modules to play with urls, the global configuration module and the sh404sef module. The global config is set to "Sef urls: YES" and "mod rewrite enabled: YES" and the sh404sef is set "url optimization: NO". My problem is, even with "Sef urls" set in the global config, my urls still don't seem to be that "user friendly" so I turn on the "Url optimization" using the sh404sef module, and I get better descriptive urls. However, the problem I inherit from doing this is that my dynamically populated chronoforms get messed up (only the chrono forms, other forms are fine); These forms are now showing up at the homepage instead of their own reserved page. Here's an example: Old form "GOOD" url: http://www.mycraftwork.com/index.php?option=com_content&view=article&id=94 New optimized "BAD" URL: http://www.mycraftwork.com/handthrown-pottery/alladin-teapot/index.php?option=com_content&view=article&id=94 Any help would be GREATLY appreciated! I can even turn the sh404sef on and off if some people are interested in seeing the issue LIVE. Thanks!! Tao Bellamine

    Read the article

  • How to configure Google Analytics experiments manually

    - by John
    I wish to run multivariate tests on an e-commerce site that run across all product pages. I will be setting and deciding the variations myself all I need to do is track the results in GA. I think may be possible (although only A/B testing is available via the GA UI): https://developers.google.com/analytics/devguides/platform/features/experiments#serving-framework EXTERNAL – You will choose variations, handle experiment optimization, and only report the chosen variation to Google Analytics. For example, this should be used by 3rd-party optimization platforms that want to integrate with Google Analytics for reporting purposes. In this case, the Google Analytics statistical engine will not run. However how do I configure this and push the data to GA in my page?

    Read the article

  • How change LOD in geometry?

    - by ChaosDev
    Im looking for simple algorithm of LOD, for change geometry vertexes and decrease frame time. Im created octree, but now I want model or terrain vertex modify algorithm,not for increase(looking on tessellation later) but for decrease. I want something like this Questions: Is same algorithm can apply either to model and terrain correctly? Indexes need to be modified ? I must use octree or simple check distance between camera and object for desired effect ? New value of indexcount for DrawIndexed function needed ? Code: //m_LOD == 10 in the beginning //m_RawVerts - array of 3d Vector filled with values from vertex buffer. void DecreaseLOD() { m_LOD--; if(m_LOD<1)m_LOD=1; RebuildGeometry(); } void IncreaseLOD() { m_LOD++; if(m_LOD>10)m_LOD=10; RebuildGeometry(); } void RebuildGeometry() { void* vertexRawData = new byte[m_VertexBufferSize]; void* indexRawData = new DWORD[m_IndexCount]; auto context = mp_D3D->mp_Context; D3D11_MAPPED_SUBRESOURCE data; ZeroMemory(&data,sizeof(D3D11_MAPPED_SUBRESOURCE)); context->Map(mp_VertexBuffer->mp_buffer,0,D3D11_MAP_READ,0,&data); memcpy(vertexRawData,data.pData,m_VertexBufferSize); context->Unmap(mp_VertexBuffer->mp_buffer,0); context->Map(mp_IndexBuffer->mp_buffer,0,D3D11_MAP_READ,0,&data); memcpy(indexRawData,data.pData,m_IndexBufferSize); context->Unmap(mp_IndexBuffer->mp_buffer,0); DWORD* dwI = (DWORD*)indexRawData; int sz = (m_VertexStride/sizeof(float));//size of vertex element //algorithm must be here. std::vector<Vector3d> vertices; int i = 0; for(int j = 0; j < m_VertexCount; j++) { float x1 = (((float*)vertexRawData)[0+i]); float y1 = (((float*)vertexRawData)[1+i]); float z1 = (((float*)vertexRawData)[2+i]); Vector3d lv = Vector3d(x1,y1,z1); //my useless attempts if(j+m_LOD+1<m_RawVerts.size()) { float v1 = VECTORHELPER::Distance(m_RawVerts[dwI[j]],m_RawVerts[dwI[j+m_LOD]]); float v2 = VECTORHELPER::Distance(m_RawVerts[dwI[j]],m_RawVerts[dwI[j+m_LOD+1]]); if(v1>v2) lv = m_RawVerts[dwI[j+1]]; else if(v2<v1) lv = m_RawVerts[dwI[j+2]]; } (((float*)vertexRawData)[0+i]) = lv.x; (((float*)vertexRawData)[1+i]) = lv.y; (((float*)vertexRawData)[2+i]) = lv.z; i+=sz;//pass others vertex format values without change } for(int j = 0; j < m_IndexCount; j++) { //indices ? } //set vertexes to device UpdateVertexes(vertexRawData,mp_VertexBuffer->getSize()); delete[] vertexRawData; delete[] indexRawData; }

    Read the article

  • Tips to Keep in Mind While Choosing an SEO Company

    If you own a website and want to have the serious profit with the help of that website then you have to find some of the ethical means by which the website can be promoted in the market and more crowds can come on your site. For this you can take advice from the people who are already into this business and can share some ideas with them. Nowadays the most common technique which is being used in the market is the SEO also known as Search Engine Optimization. There are many people on the web world who are familiar with this term but don't have a full idea. For such people there are many Search Engine Optimization Companies which are assisting the people.

    Read the article

  • What does the SEO market look like? [closed]

    - by TheEconomist
    I'm an economist (aspiring economist, if we are being technical) and my curiousity has been recently piqued by search engine optimization services. I had the following questions. 1) What industries most widely use search engine optimization services? 2) Is it small business or big firms that use SEO services typically? 3) What is SEO pricing typically dependent upon and how much is it on average? 4) Do SEO services really increase traffic? Is there a dataset I can get a hold of for this sort of thing? I am not looking for answers to the questions necessarily. Although answers would help, a nudge in the right direction is more than sufficient, and greatly appreciated. Thank you!

    Read the article

  • Boosting my GA with Neural Networks and/or Reinforcement Learning

    - by AlexT
    As I have mentioned in previous questions I am writing a maze solving application to help me learn about more theoretical CS subjects, after some trouble I've got a Genetic Algorithm working that can evolve a set of rules (handled by boolean values) in order to find a good solution through a maze. That being said, the GA alone is okay, but I'd like to beef it up with a Neural Network, even though I have no real working knowledge of Neural Networks (no formal theoretical CS education). After doing a bit of reading on the subject I found that a Neural Network could be used to train a genome in order to improve results. Let's say I have a genome (group of genes), such as 1 0 0 1 0 1 0 1 0 1 1 1 0 0... How could I use a Neural Network (I'm assuming MLP?) to train and improve my genome? In addition to this as I know nothing about Neural Networks I've been looking into implementing some form of Reinforcement Learning, using my maze matrix (2 dimensional array), although I'm a bit stuck on what the following algorithm wants from me: (from http://people.revoledu.com/kardi/tutorial/ReinforcementLearning/Q-Learning-Algorithm.htm) 1. Set parameter , and environment reward matrix R 2. Initialize matrix Q as zero matrix 3. For each episode: * Select random initial state * Do while not reach goal state o Select one among all possible actions for the current state o Using this possible action, consider to go to the next state o Get maximum Q value of this next state based on all possible actions o Compute o Set the next state as the current state End Do End For The big problem for me is implementing a reward matrix R and what a Q matrix exactly is, and getting the Q value. I use a multi-dimensional array for my maze and enum states for every move. How would this be used in a Q-Learning algorithm? If someone could help out by explaining what I would need to do to implement the following, preferably in Java although C# would be nice too, possibly with some source code examples it'd be appreciated.

    Read the article

  • How do I use WS-Security with WCF?

    - by Jonathan Allen
    Below is the style of header I need to create. I am expected to use either a public/private key or a SSL style certificate. I don't know for certain, but I think my counter-party is using some form of Java. <soap-env:Header> <wsse:Security xmlns:wsse="http://schemas.xmlsoap.org/ws/2002/04/secext"> <ds:Signature xmlns:ds="http://www.w3.org/2000/09/xmldsig#"> <ds:SignedInfo xmlns:ds="http://www.w3.org/2000/09/xmldsig#"> <ds:CanonicalizationMethod Algorithm="http://www.w3.org/2001/10/xml-exc-c14n#" /> <ds:SignatureMethod Algorithm="http://www.w3.org/2000/09/xmldsig#rsa-sha1" /> <ds:Reference URI="#secinfo"> <ds:DigestMethod Algorithm="http://www.w3.org/2000/09/xmldsig#sha1" /> <ds:DigestValue>xxxxxxxxxxxxx</ds:DigestValue> <ds:Transforms> <ds:Transform Algorithm="http://www.w3.org/TR/1999/REC-xpath-19991116"> <ds:XPath>//*[@id='secinfo']/child::*/text()</ds:XPath> </ds:Transform> </ds:Transforms> </ds:Reference> </ds:SignedInfo> <ds:SignatureValue>xxxxxxxxxxds:SignatureValue> <ds:KeyInfo> <ds:KeyName>xxxxxxx</ds:KeyName> </ds:KeyInfo> </ds:Signature> <t:UsernameToken xmlns:t="http://schemas.xmlsoap.org/ws/2002/04/secext" id="secinfo"> <t:UserInfo>USER=xxxx;CORR=xxxx;TIMESTAMP=201003161916</t:UserInfo> </t:UsernameToken> </wsse:Security> </soap-env:Header>

    Read the article

  • Map large integer to a phrase

    - by Alexander Gladysh
    I have a large and "unique" integer (actually a SHA1 hash). I want (for no other reason than to have fun) to find an algorithm to convert that SHA1 hash to a (pseudo-)English phrase. The conversion should be reversible (i.e., knowing the algorithm, one must be able to convert the phrase back to SHA1 hash.) The possible usage of the generated phrase: the human readable version of Git commit ID, like a motto for a given program version (which is built from that commit). (As I said, this is "for fun". I don't claim that this is very practical — or be much more readable than the SHA1 itself.) A better algorithm would produce shorter, more natural-looking, more unique phrases. The phrase need not make sense. I would even settle for a whole paragraph of nonsense. (Though quality — englishness — of a paragraph should probably be better than for a mere phrase.) A variation: it is OK if I will be able to work only with a part of hash. Say, first six digits is OK. Possible approach: In the past I've attempted to build a probability table (of words), and generate phrases as Markov chains, seeding the generator (picking branches from probability tree), according to the bits I read from the SHA. This was not very successful, the resulting phrases were too long and ugly. I'm not sure if this was a bug, or the general flaw in the algorithm, since I had to abandon it early enough. Now I'm thinking about attempting to solve the problem once again. Any advice on how to approach this? Do you think Markov chain approach can work here? Something else?

    Read the article

  • The Immerman-Szelepcsenyi Theorem

    - by Daniel Lorch
    In the Immerman-Szelepcsenyi Theorem, two algorithms are specified that use non-determinisim. There is a rather lengthy algorithm using "inductive counting", which determines the number of reachable configurations for a given non-deterministic turing machine. The algorithm looks like this: Let m_{i+1}=0 For all configurations C Let b=0, r=0 For all configurations D Guess a path from I to D in at most i steps If found Let r=r+1 If D=C or D goes to C in 1 step Let b=1 If r<m_i halt and reject Let m_{i+1}=m_{i+1}+b I is the starting configuration. m_i is the number of configurations reachable from the starting configuration in i steps. This algorithm only calculates the "next step", i.e. m_i+1 from m_i. This seems pretty reasonable, but since we have nondeterminisim, why don't we just write: Let m_i = 0 For all configurations C Guess a path from I to C in at most i steps If found m_i = m_i + 1 What is wrong with this algorithm? I am using nondeterminism to guess a path from I to C, and I verify reachability I am iterating through the list of ALL configurations, so I am sure to not miss any configuration I respect space bounds I can generate a certificate (the list of reachable configs) I believe I have a misunderstanding of the "power" of non-determinisim, but I can't figure out where to look next. I am stuck on this for quite a while and I would really appreciate any help.

    Read the article

  • x509 certificate Information

    - by sid
    Certificate: Data: Version: 3 (0x2) Serial Number: 95 (0x5f) Signature Algorithm: sha1WithRSAEncryption Issuer: C=, O=, CN= Validity Not Before: Apr 22 16:42:11 2008 GMT Not After : Apr 22 16:42:11 2009 GMT Subject: C=, O=, CN=, L=, ST= Subject Public Key Info: Public Key Algorithm: rsaEncryption RSA Public Key: (1024 bit) Modulus (1024 bit): ... ... ... Exponent: 65537 (0x10001) X509v3 extensions: X509v3 Key Usage: critical Digital Signature, Key Encipherment X509v3 Extended Key Usage: critical Code Signing X509v3 Authority Key Identifier: keyid: ... Signature Algorithm: sha1WithRSAEncryption a9:55:56:9b:9e:60:7a:57:fd:7:6b:1e:c0:79:1c:50:62:8f: ... ... -----BEGIN CERTIFICATE----- ... ... ... -----END CERTIFICATE----- In This Certificate, Which is the public key? is Modulus? what does the Signature Algorithm, a9:55:56:... represent (is it message digest)? And what is between -----BEGIN CERTIFICATE----- & -----END CERTIFICATE-----, is That the whole certificate? As I am novice, little bit confusing between the message digest and public key? Thanks in Advance-opensid

    Read the article

  • slowAES encryption and java descryption

    - by amnon
    Hi , I've tried to implement the same steps as discussed in AES .NET but with no success , i can't seem to get java and slowAes to play toghter ... attached is my code i'm sorry i can't add more this is my first time trying to deal with encryption would appreciate any help private static final String ALGORITHM = "AES"; private static final byte[] keyValue = getKeyBytes("12345678901234567890123456789012"); private static final byte[] INIT_VECTOR = new byte[16]; private static IvParameterSpec ivSpec = new IvParameterSpec(INIT_VECTOR); public static void main(String[] args) throws Exception { String encoded = encrypt("watson?"); System.out.println(encoded); } private static Key generateKey() throws Exception { Key key = new SecretKeySpec(keyValue, ALGORITHM); // SecretKeyFactory keyFactory = SecretKeyFactory.getInstance(ALGORITHM); // key = keyFactory.generateSecret(new DESKeySpec(keyValue)); return key; } private static byte[] getKeyBytes(String key) { byte[] hash = DigestUtils.sha(key); byte[] saltedHash = new byte[16]; System.arraycopy(hash, 0, saltedHash, 0, 16); return saltedHash; } public static String encrypt(String valueToEnc) throws Exception { Key key = generateKey(); Cipher c = Cipher.getInstance("AES/CBC/PKCS5Padding"); c.init(Cipher.ENCRYPT_MODE, key,ivSpec); byte[] encValue = c.doFinal(valueToEnc.getBytes()); String encryptedValue = new BASE64Encoder().encode(encValue); return encryptedValue; } public static String decrypt(String encryptedValue) throws Exception { Key key = generateKey(); Cipher c = Cipher.getInstance(ALGORITHM); c.init(Cipher.DECRYPT_MODE, key); byte[] decordedValue = new BASE64Decoder().decodeBuffer(encryptedValue); byte[] decValue = c.doFinal(decordedValue); String decryptedValue = new String(decValue); return decryptedValue; } the bytes returned are different thanks in advance .

    Read the article

  • C++ Namespaces & templates question

    - by Kotti
    Hi! I have some functions that can be grouped together, but don't belong to some object / entity and therefore can't be treated as methods. So, basically in this situation I would create a new namespace and put the definitions in a header file, the implementation in cpp file. Also (if needed) I would create an anonymous namespace in that cpp file and put all additional functions that don't have to be exposed / included to my namespace's interface there. See the code below (probably not the best example and could be done better with another program architecture, but I just can't think of a better sample...) Sample code (header) namespace algorithm { void HandleCollision(Object* object1, Object* object2); } Sample code (cpp) #include "header" // Anonymous namespace that wraps // routines that are used inside 'algorithm' methods // but don't have to be exposed namespace { void RefractObject(Object* object1) { // Do something with that object // (...) } } namespace algorithm { void HandleCollision(Object* object1, Object* object2) { if (...) RefractObject(object1); } } So far so good. I guess this is a good way to manage my code, but I don't know what should I do if I have some template-based functions and want to do basically the same. If I'm using templates, I have to put all my code in the header file. Ok, but how should I conceal some implementation details then? Like, I want to hide RefractObject function from my interface, but I can't simply remove it's declaration (just because I have all my code in a header file)... The only approach I came up with was something like: Sample code (header) namespace algorithm { // Is still exposed as a part of interface! namespace impl { template <typename T> void RefractObject(T* object1) { // Do something with that object // (...) } } template <typename T, typename Y> void HandleCollision(T* object1, Y* object2) { impl::RefractObject(object1); // Another stuff } } Any ideas how to make this better in terms of code designing?

    Read the article

< Previous Page | 181 182 183 184 185 186 187 188 189 190 191 192  | Next Page >