From wiki.eser.org

Contents 1 About ESer 1.1 what does ESer do? 1.2 What does ESer stand for? 1.3 Why did you create ESer? 1.4 How does ESer know to recommend a document and/or keyword that doesn't contain my search terms? 1.5 Where can I download the ESer source code? 1.6 Can ESer suggest keywords and/or documents on my private dataset? 2 ESer Feedback 3 erlang 3.1 installing erlang 3.2 erlang hello world 3.3 erlang concurrent hello world 3.4 good erlang resources 3.4.1 websites 3.4.2 videos 3.4.3 cool erlang articles 3.4.4 massively parallel erlang database systems 4 How to deal with Ethernet latency 4.1 Cool solid-state hardware solutions for avoiding ethernet latency in high-throughput computing applications with large data sets 4.1.1 Fusion IO NAND Flash PCI-E Card 4.1.1.1 stats 4.1.2 using video memory cards as swap space on Linux 5 wikipedia-specific search engines and datamining tools 6 Machine learning 7 Open access public datasets for machine learning and datamining applications 8 Business and Law 9 Cluster farm infrastructure 10 all about smb and cifs 11 wiki's 12 HCI and Usability 13 Virtual Appliances for Software Deployment

[edit] About ESer

[edit] what does ESer do?

• short-version: ESer recommends interesting keywords, ideas, and documents
• long-version: ESer is a contextual k-nn similarity search engine using probabilistic latent semantic indexing, latent dirichlet allocation, compressed sensing, and stream computing on erlang.

[edit] What does ESer stand for?

• Exponential Systolic Epistemic Recursion

[edit] Why did you create ESer?

• Because we invented what we think is a really cool algorithm; then we tried to find a useful application for it.
  • The Eser algorithm was originally invented to improve autonomous mobile robot navigation and localization.
    • the eser.org website is simply a demo of how the ESer algorithm can be applied to solve problems outside of robotics.
• If you can think of a better application for this technology, let us know

[edit] How does ESer know to recommend a document and/or keyword that doesn't contain my search terms?

• ESer uses a variation of Probabilistic latent semantic analysis called Latent Dirichlet allocation

[edit] Where can I download the ESer source code?

• Keep an eye on the ESer homepage for updates
  • We're currently improving our algorithm based on user-feedback before releasing the ESer algorithm as a stand-alone software package.
  • The source code for the ESer algorithm will be released under GPL v2
  • Concurrent with the release of the source code, we will be releasing a paper that discusses the ESer algorithm

[edit] Can ESer suggest keywords and/or documents on my private dataset?

• Probably. Email the following address for more information: support [AT] eser [D O T] org

[edit] ESer Feedback

• Top User Requests for Eser
• Todo List

[edit] erlang

[edit] installing erlang

• to install erlang on Debian Linux or Ubuntu Linux, do the following:

apt-get -y install erlang erlang-examples erlang-manpages erlang-mode erlang-src

• if you use emacs, you're golden: this will automatically install a nifty erlang-editing mode for emacs

[edit] erlang hello world

$ erl -noshell -eval 'io:fwrite("Hello, world!\n"), init:stop().'
Hello, world!

[edit] erlang concurrent hello world

• source: Andre Pang talk on concurrency with erlang
• create a file named hello_concurrent.erl with the following contents:

-module( hello_concurrent ).
-export( [ receiver/0, giver/1, start/0 ] ).

receiver() ->
    receive
        diediedie ->
            ok;
        { name, Name } ->
             io:format( "hello, ~s~n", [ Name ] ),
             receiver()
    end.

giver( ReceiverPid ) ->
    ReceiverPid ! { name, "World" },
    ReceiverPid ! { name, "Everyone" },
    ReceiverPid ! diediedie.

start() ->
    ReceiverPid = spawn( hello_concurrent, receiver, [] ),
    spawn( hello_concurrent, giver, [ ReceiverPid ] ),
    start_finished.

• enter the erlang shell by typing erl
• run the program from inside the erlang shell by typing the following:

c(hello_concurrent).
hello_concurrent:start().

[edit] good erlang resources

• Programming Erlang: Software for a Concurrent World by Joe Armstrong
  • written by the implementor of the first version of erlang -- the definitive introductory text to erlang

[edit] websites

• the erlang community website -- howtos, tutorials, erlang cookbook, articles, blog posts, mailing lists
  • actively maintained and updated website
• the erlang FAQ

[edit] videos

• july 2007 google techtalk on erlang -- provides general overview and introduction; link forwards to start of sample code
• google tech talk on Kilim -- attempt to retrofit Java for erlang-like lightweight threads

[edit] cool erlang articles

• well-written article on implementing simple feed-forward ANN with backprop in erlang
  • Each process is a perceptron, and acts as an autonomous agent that only interacts with other processes it comes into contact with--in this case, only other perceptrons it's connected to.
• how to interface perlang with code from other languages
  • c, c++, etc...
  • you can re-use your perl code in erlang projects by using unix pipes

[edit] massively parallel erlang database systems

1. CouchDB

[edit] How to deal with Ethernet latency

• for a good overview of why this is such a hard problem, see: Andy Bechtolsheim's stanford excellent ee380 2006 talk on high-performance compute-cluster interconnects which is available in streaming video for viewing here.
• there are two approaches to addressing this problem
  1. the only power-efficient and cost-efficient option is to size the hardware and the dataset to keep all compute I/O local to the compute node. This creates a need for high-density and low-latency solid-state storage technologies to extend the cpu_registers->multilevel_cache->main_memory latency hierarchy
  2. the more common approach is to hide ethernet latency by using offline batch-computing operations.see: Google MapReduce: Simplified Data Processing on Large Clusters

[edit] Cool solid-state hardware solutions for avoiding ethernet latency in high-throughput computing applications with large data sets

[edit] Fusion IO NAND Flash PCI-E Card

• the Fusion IO iodrive is a solid-state NAND flash storage device with a capacity of up to 320GB, random access latencies of 40 microseconds for writes and 50 microseconds for reads

[edit] stats

• 40 microsecond write latency
• 50 microsecond read latency
• capacity: up to 320GB
• sustained throughput: 2.8GB/s read 2.4GB/s write (RAID-0 configuration with 4 cards)
• power consumption: 6 watts (per card)

[edit] using video memory cards as swap space on Linux

• idea: on a motherboard with 4 x pci-e slots, you could populate each slot with a 1GB+ video card and use the linux kernels Memory_Technology_Device support to map the video memory in PCI space into "uncached system RAM"
  • downside: high-end video cards rarely offer more than a few GB of memory
  • howto do this: Gentoo Wiki TIP: using memory on video card as swap

[edit] wikipedia-specific search engines and datamining tools

• ESer.org -- semantic search engine that is currently text-mining wikipedia
• Wikiscanner -- list anonymous wikipedia edits by reverse IP address lokup

[edit] Machine learning

• duplicate insensitive cardinality estimation in distributed databases
• duplicate insensitive uniform random sampling of distinct values
• approximate duplicate document detection
• machine learning resources
• ngram minhashing
• POS taggers and phrase chunkers

[edit] Open access public datasets for machine learning and datamining applications

• SEC EDGAR database
• Wikipedia database dump
• Wiktionary dictionary database dump
• Library of Congress semantic database dump (with links)
• JURIS database -- 2-million-page archive of federal case law.
  • 900 Mbyte tarball

[edit] Business and Law

• GPL v2 notes
• How the customer benefits from leasing
• On Open Source
• Tech Industry Trends

[edit] Cluster farm infrastructure

• How To Use Multiple Geographically Distributed Datacenters and POP's
• Memcached notes

[edit] all about smb and cifs

• how to write an intranet windows file shares crawler
• SMB/CIFS Robots

[edit] wiki's

• what are wiki's good for?
• mediawiki administration
• How to use ESER to text-mine your wiki
• How to build a password-protected wiki server from scratch for internal use by your organization

[edit] HCI and Usability

• HCI and Usability web resources

[edit] Virtual Appliances for Software Deployment

• How to create VMWare virtual machine appliances