Dustin Turner

Search the log and builds

Search posts, builds, and glossary terms.

Learning log · Updated as I go · Rev A

2,101 topics stand between an operations seat and an engineering one. I’ve checked 17.

Two checklists — machine learning and robotics — mapping everything I’d need to learn to do this work for real. Every box I check has a published explainer behind it, because a checked box with nothing to show for it isn’t evidence of anything.

A site about the distance between a demo and a deployment has no business hiding its own denominator. So this is the honest version: the size of the territory, my position in it, and the artifacts that prove I was there.

Topics mapped
2,101
Boxes checked
17
Explainers published
16
Percent of the map
0.8%
The checklists · Live
Explainers · 16 published

One box checked, one explainer shipped.

Interactive, plain-English, and written for the person I was three months ago. If I can’t explain it clearly enough to publish, I don’t get to call it learned.

Machine learningScalars, vectors, matrices, tensorsthe four object types and how dimensions are written and readRead the explainer ↗Machine learningVector addition and scalar multiplicationthe two operations that define a vector spaceRead the explainer ↗Machine learningDot product / inner productmeasures alignment between vectors; the single most common operation in MLRead the explainer ↗Machine learningEigenvalues and eigenvectorsdirections preserved by a transformation, scaled by a factorRead the explainer ↗Machine learningConditional probabilityprobability given that something else is knownRead the explainer ↗Machine learningLaplace distributionheavier tails than Gaussian, the prior corresponding to L1 regularizationRead the explainer ↗Machine learningMarkov chains and stationary distributionsmemoryless stochastic processesRead the explainer ↗Machine learningGradient descentthe fundamental iterative algorithmRead the explainer ↗Machine learningpandas Series and DataFramelabeled tabular dataRead the explainer ↗Machine learningLinear regressionpredicting a continuous target as a weighted sum of featuresRead the explainer ↗Machine learningLogistic regressionpredicting binary class probability via the sigmoidRead the explainer ↗Machine learningRandom forestsbagged trees with random feature subsets at each splitRead the explainer ↗Machine learningWhat NLP isturning text into computable representations and backRead the explainer ↗Machine learningBayesian inferenceupdating beliefs with evidenceRead the explainer ↗RoboticsVectors and vector spacesthe basic object; positions, velocities, forces, and torques are all vectorsRead the explainer ↗RoboticsDot productprojection and alignment; work is force dotted with displacementRead the explainer ↗
The method · Why it's public
  1. 01

    The map comes first

    Both checklists were written before I started, so the denominator can't move to flatter the numerator.

  2. 02

    A box is only checked when something ships

    An explainer, a build, or a written teardown. Reading about a topic is not the same as understanding it, and only one of those leaves evidence.

  3. 03

    The gaps stay visible

    Nothing is hidden. You can see every topic I haven't touched, which is nearly all of them.

  4. 04

    The floor is the lab

    I work around industrial automation every shift. When the textbook and the deployment disagree, that disagreement is the post.

The ML checklist →The robotics checklist →