bkoshelev / problem-solution-markdown-templates

Markdown Templates to build a task-targeted knowledge base for work in Machine Learning / Data Science

Problem-solution Markdown templates

Markdown Templates for a task-targeted personal knowledge database.

Goal

These templates are to function as a practical problem-solution lookup with each solution including

• use case
• assumptions
• example
• code example
• references (internal links)
• library refs

Organization and Note Types (goal of quick useful reference)

Provided templates

• problem domain type (structure note)
• problem type
• solution type
• object / concept type (structure or base type)

General workflow:

• find note corresponding to the type/topic the problem with which you are struggling resides (search topics in UU-problem-type or specific note with type:structure, ie U-problem-type note)
• within U-problem-type find specific problem-note corresponding to the specific difficulty you are facing: this will be type:problem (base level)
• to each problem-note will correspond blocks representing contexts or assumptions. This dictates the linked solution note: each with ways to solve the relevant problem and related references (concept, implementation, reference texts etc)

** Cheat Sheet on Note Types:**

• top level structure note for general problem topics are of type:structure/UU and have titles of form UU-title. (example "UU-model-generalization")
• More specific sub-type of the given problem are of type:structure/U with titles of form U-title ("U-overfitting”)
• lists to go through in creating notes which like a container briefly hold thoughts are “buffer” notes: §-title

** The following are base level notes:**

• type:problem
• type:solution notes ("L1 regularization for overfitting")
• type:object ("penalty term") are reference terms and objects.

** front matter cheat sheet:**

I have valid (obsidian) Yaml followed by valid json for extra metadata parsing.

I like the ability to always create online flash cards just with :: so flashcards is always tagged in case I use it.

—— alias:

• title
• id (YYMMDDHHmm)
• a1
• a2

tags: for me tags are used for obsidian plug-in hot-keys

• todo
• flashcards ——

{ json: “extra metadata to parse, K2: “…”, … }

If you have a Job / Training focus

1. Taylor the problem types to what you exoect to encounter at your job.

2. focus on what might take you too much time to hunt for solutions in real-time.

3. be sure examples are well annotated

4. include references to relevant functions in associated libraries

5. Flow: problem type -> problem -> solution under assumptions -> examples

I hope these are helpful