Loading presentation...

Present Remotely

Send the link below via email or IM

Copy

Present to your audience

Start remote presentation

  • Invited audience members will follow you as you navigate and present
  • People invited to a presentation do not need a Prezi account
  • This link expires 10 minutes after you close the presentation
  • A maximum of 30 users can follow your presentation
  • Learn more about this feature in our knowledge base article

Do you really want to delete this prezi?

Neither you, nor the coeditors you shared it with will be able to recover it again.

DeleteCancel

Make your likes visible on Facebook?

Connect your Facebook account to Prezi and let your likes appear on your timeline.
You can change this under Settings & Account at any time.

No, thanks

libalexandria Architecture

RCOS Debut
by

Tor Hagemann

on 2 April 2013

Comments (0)

Please log in to add your comment.

Report abuse

Transcript of libalexandria Architecture

A Closer Look... Up-close and personal libalexandria High-Level
(Java) API Goal: Complete Javadoc Swagger Low-Level C/
Fortran API (linked via JNI) Building an
Efficient Bridge Custom Back-End Our tagline:
"What's Your Intent?" An RCOS
Spring 2013 Project Into The Wild Blue Yonder Data Personal Management The Cloud Customer
Relationship Queries Recommendations Processing Cluster Work
Units Better
User Experiences SWIG:
Simplified
Wrapper and
Interface
Generator C/C++
Libraries Ruby Python Java Perl, PHP, Tcl,
C#, Lisp, D, Go, Lua,
OCAML, Octave, R... Android API Java "Lannister" Your
App Learning Model Supervised Reinforcement Unsupervised Labels Feature
Selection Logging Profiling Linear Regression
Logistic Regression
Support Vector Machine
(SVM w/ Kernel) Clustering K-Means Signal Separation Neural Networks Hierarchical
Temporal Memory Find a hard problem! Collect Samples! Training Day Applying What
You've Learned Tune Behavior
to Meet Goals Automatic
Model-Selection
Framework * Identify Your Domain
* Program User-Experience
* Request Android ML Service To the high-level API... * Construct Model
* Extract Features
* Label Samples
* Scale Data
* Prepare to stream! * Pipe-in known solutions
* Model extrapolates patterns
* Trained to classify unknowns
* Tunes automatically!
* profiling is built in
* models are parameter-aware
* e.g. regularization * Data comes back labeled
* Define resulting behavior
* Completely please your users
* Identify trends/issues
* Make interesting suggestions * Not fast enough? (simplify!)
* Wrong variables? (swap them!)
* Need additional resources?
* Tweak your prototypes quickly
to find optimality!
(* finally: succeed
ridiculously) Leveraging:
Machine Learning
Onion-Skin API
Interactive modeling * libalexandria starts daemon
* communicate via "intents"
(think OpenCV) * Make the world your playground
* Interact fluidly with developers
* Compute on your terms Sound
BIG? SGEMM,
DGEMM,
CGEMM,
ZGEMM SSYMM,
DSYMM,
CSYMM,
ZSYMM STRMM,
DTRMM,
CTRMM,
ZTRMM BLAS lvl3
(Matrices) BLAS lvl1/2 Adds scalar routines like: SSYRK,
DSYRK,
CSYRK,
ZSYRK DOT, NORM, etc.
ROTATE, MIN/MAX
TRANSPOSE,
SUM/DIFF Vectorized Multiplication High-level API That's only
the tip of the
iceberg! And beyond! Efficiency Snobs! Scripting! User
Experience! Web
Apps! Within the "Libraries of Alexandria" Buzzwords! The
Aquaduct Data-
Driven
Pipeline (for computation) BLAS Library
(ATLAS/similar
optimized
linear-algebra
routines Fortran 2003
ISO_C_BINDING
w/ modular
organization "Glue" Code
(in ANSI C)
[libkolla]
[libgefyra] Toy
Apps
(PoC) Inventory (shhhhhh...) by:
Tor E Hagemann
and you?
Full transcript