Commit 54d418bf authored by Robert Ricci's avatar Robert Ricci

Snapshot: Made it through mistake #14

parent 77ac7535
......@@ -20,11 +20,43 @@
\1 Big thing one
\2 Little thing A
\2 Little thing B
\1 Big thing two
\1 Left over from last time
\2 Who still doesn't have the textbook?
\2 To post assignments on canvas or not?
\1 Today is all about mistakes
\1 Mistakes from the book:
\2 No goals: identify problem (metrics, workloads, and methodology)
\2 Biased goals: prove my system better than theirs
\2 Unsystematic approach: If you miss things, your conclusion may be wrong
\2 Analysis without understanding the problem: ``A problem well stated is
half solved'. Especially a problem when applying solutions to areas
outside expertise
\2 Incorrect performance metrics: Metrics that solve the problem, comparing
apples to apples (eg. MIPS for CISC vs. RISC) Often chose easy rather
than relevant
\2 Unrepresentative workload: Representative of actual usage
\2 Wrong technique: Using the hammer that you have. Are you capturing all
important parts of the environment?
\2 Overlooking important parameters: Choice of parameters matters a lot
\2 Ignoring significant factors: factors are the parameters you vary.
Scientific method; control vs. variables. Sensitivity analysis:
something we rarely do!
\2 Inappropriate experiment design: how many experiments, how many
trials, parameter choice
\2 Inappropriate levels of detail: ``Forest for the trees''
\2 No analysis: have to tell the reader what they see
\2 Erroneous analysis: base analysis on bogus techniques such as
average of ratios or bad experiments
\2 No sensitivity analysis: Analysis gives you evidence, not facts.
What would change if setting changes?
\1 For next time
\2 Read 2.2
\2 HW \#2 due before class starts
\2 HW \#3 will be handed out by class time
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment