Engineering Probability Class 21 Thu 2018-04-05

Source

1   Parallel computer access

As I mentioned Monday, parallel.ecse.rpi.edu is available. It has a dual 14-core (56 hyperthread) 2GHz Intel Xeon CPU, Intel Xeon Phi coprocessor with 60 cores running 240 threads, Nvidia GeForce GTX 1080 GPU with 2560 CUDA cores, and 256GB main memory.

Parallel SW includes CUDA, Thrust, OpenMP, TBB.

Using it for private business is against RPI policy.

It is slower than an IBM Blue Gene, but it cost only $10K total.

I manage it; ask me for an account.

Many datasets that are considered to be big data will fit into main memory; MPI etc are not required.

2   Final exam

will be 80 minutes like the first 2 exams.

3   Material from text

  1. Example 4.33, page 177.

  2. Example 4.36, page 180, Amplitude Samples of a Sinusoidal Waveform

  3. Equation 5.32 on page 259.

  4. Example 5.27 on page 260. Uncorrelated but Dependent Random Variables.

  5. Example 5.29 on page 263. Loaded Dice

  6. Example 5.30 on page 263. Number of Defects in a Region; Random Splitting of Poisson Counts

  7. Example 5.31 on page 264. This is a noisy comm channel, now with Gaussian (normal) noise. This is a more realistic version of the earlier example with uniform noise. The application problems are:

    1. what input signal to infer from each output,
    2. how accurate is this, and
    3. what cutoff minimizes this?

    In the real world there are several ways you could reduce that error:

    1. Increase the transmitted signal,
    2. Reduce the noise,
    3. Retransmit several times and vote.
    4. Handshake: Include a checksum and ask for retransmission if it fails.
    5. Instead of just deciding X=+1 or X=-1 depending on Y, have a 3rd decision, i.e., uncertain if $|Y|<0.5$, and ask for retransmission in that case.

    This is relevant to solving Problem 5.3, page 288.