From d959ff9203abf8d6b6297c12c465433e3b4ce0f3 Mon Sep 17 00:00:00 2001 From: Oliver Kennedy Date: Thu, 6 Dec 2018 22:45:45 -0500 Subject: [PATCH] Turek talk --- slides/talks/2018-6-Matt_Turek/index.html | 98 + .../talks/2018-6-Matt_Turek/ml-pipeline.svg | 3020 +++++++++++ .../talks/2018-6-Matt_Turek/tag-pipeline.svg | 4717 +++++++++++++++++ 3 files changed, 7835 insertions(+) create mode 100644 slides/talks/2018-6-Matt_Turek/index.html create mode 100644 slides/talks/2018-6-Matt_Turek/ml-pipeline.svg create mode 100644 slides/talks/2018-6-Matt_Turek/tag-pipeline.svg diff --git a/slides/talks/2018-6-Matt_Turek/index.html b/slides/talks/2018-6-Matt_Turek/index.html new file mode 100644 index 00000000..2c4a317e --- /dev/null +++ b/slides/talks/2018-6-Matt_Turek/index.html @@ -0,0 +1,98 @@ + + + + + + + reveal.js + + + + + + + + + + + +
+ +
+

Quality-Aware Machine Learning

+

Oliver Kennedy

+

Jaroslaw Zola

+

Matthew Knepley

+
+ +
+ +
+ +
+
+

Fixing Data is Expensive

+

(or impossible)

+
+ +
+

Re-using already fixed data is dangerous.

+

(the "right" fix depends on use case)

+
+
+ +
+
+

Idea: Track Errors

+ +

Incomplete Databases store possibilities, not just certainties.

+
+ +
+ +
+ +
+

Goals

+
    +
  • Statistically rigorous techniques for training classifiers, neural networks on incomplete databases.
    • + +
  • Models incorporating incompleteness information. +
    "I didn't have enough training data" should be an allowed prediction.
    +
    • + +
  • Incompleteness as an assist for model debugging. +
    Which errors have the biggest impact on a prediction?
    + Which errors best explain an incorrect prediction?
    +
    • +
+ +
+ +
+ + + + + + + diff --git a/slides/talks/2018-6-Matt_Turek/ml-pipeline.svg b/slides/talks/2018-6-Matt_Turek/ml-pipeline.svg new file mode 100644 index 00000000..65ca3ecb --- /dev/null +++ b/slides/talks/2018-6-Matt_Turek/ml-pipeline.svg @@ -0,0 +1,3020 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + image/svg+xml + + + + + + + + + + + + + + + + 010111010011101011101001011010101001010010101 + + Data + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + Explore + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + Model + + + + + + + + + + + + + + REPORT + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + Predict + + + + + + + ID Error + + + + + + + + + Pick Fix + + + + + + Why is it an error?What are my options?What are pros/cons?Does a fix exist? + + + + + + + + + + diff --git a/slides/talks/2018-6-Matt_Turek/tag-pipeline.svg b/slides/talks/2018-6-Matt_Turek/tag-pipeline.svg new file mode 100644 index 00000000..27dc4c19 --- /dev/null +++ b/slides/talks/2018-6-Matt_Turek/tag-pipeline.svg @@ -0,0 +1,4717 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + image/svg+xml + + + + + + + + + + + + + + + 010111010011101011101001011010101001010010101 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + REPORT + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + ID Error + + + + + + + + + Tag Error + + + Faster + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + REPORT + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + Reusable + + + + + More Reliable + Easier Debugging + +