edcote

Chapter 1

• "Iron law": 1/Perf = time/program = instructions/program (cycle count) * cycles/instruction (CPI) * time/cycle (cycle time)
• "Amdahl's law" = speedup = 1 / time = 1 / ((1-f)+(f/N))
  • speedup is limited by sequential bottlenec

Chapter 2

• Three possible data dependences between two instructions, true (RAW), anti (WAR), and output (WAW). Also applies to memory data dependencies (not applicable in simple five stage pipeline).
• There is also control dependencies.

edcote

Notes for 'Scala for Machine Learning, P. Nicolas" url

• Chapter 1

Critical to understand the different classes of ML algorithms and to select the ones that are relevant to the domain.

ML problems are categorized as classification, prediction, optimization, and regression.

• Classification is to extract knowledge for historical data. For example, a classifier can be built to identify a disease from a set of symptoms.

edcote

Installation

TBD

Basic Concepts

See: https://www.analyticsvidhya.com/blog/2017/01/scala/

• Lazy operation: operations that do not execute until we require results
• Spark Context: holds a context with Spark cluster manager

edcote

CM1K Chip

Each neuron consists of SRAM and a small programmable logic unit. The logic is prewired to run certrain types of algorithms. Neurons are interconnected using a small bidir bus.

Released in 2007. Follow up to IBM ZISC chip. ZISC refers to an architecture based solely on pattern matching and the abscense of micro-instructions. A single ZISC036 holds 36 neurons to implement an RBF network trained with the RCE (or ROI) algorithm.

ZISC employs Radial Basis Function (RBF) and K-Nearest Neighbor (KNN) algorithms. ZISC approach is a specialized but cheap chip to do one thing very quickly.

• RBF: Real-valued function whose value depends only on the distance from the origin. Used as a kernel in support vector classification. Can be interpreted as a simple single-layer type of ANN.
• KNN: Stores all avalable cases and classifies new cases based on a similarly measure (e.g. distance function). Used in statistical estimation and pattern recognition.

edcote

• Different technique needed for GUI and non-GUI login:

https://superuser.com/questions/1200539/cannot-increase-open-file-limit-past-4096-ubuntu

• Contents of /etc/security/limits.conf. For non-GUI login:

# FIXME: is * and root required?
@plus            soft    nofile          131072
@plus hard nofile 131072

edcote

Installation

Download tool from here. Unzip, then launch: bin/zeppelin-daemon.sh start

Then, browse to http://localhost:9009/#/

Here is an example Systemd service file:

$ cat /etc/systemd/system/zeppelin.service 

[Unit]

edcote

MongoDB

MongoDB stores data in flexible, JSON-like documents. The document model maps to object in your application code, making data easy to work with. MongoDB is a distributed database at its core.

Here. Be careful, need 3.4 or later.are notes on how to install MongoDB on RHEL7.

Spark and MongoDB

MongoDB is a document store (and essentially a database). Spark is a computing engine and not a store. When used together, Spark jobs can be executed directly on operational data sitting in MongoDB.

edcote

Why CMake? Why? It is only me or do I find CMake to be convoluted and non intuitive?

• Debug vs. Release:

Technically, this is all that is needed:

cmake -DCMAKE_BUILD_TYPE=Release ..
cmake -DCMAKE_BUILD_TYPE=Debug ..

edcote

Scala Notes

Variables

Immutable and mutable

Scala has two kinds of variables, vals and vars:

• val is immutable, cannot be reassigned
• var is mutable, can be reassigned

edcote

Summary

• Networks must be hierarchically composable. Topology must form DAG.
• MOESI-equivalent protocol

Channels

• Channel A : M-S : Request
• Channel D : S-M : Data response
• Channel B : M-S : Snoop (Request that an operation to be performance at an address cached by a master, accessing, or writing back that cached data)