http://www.se-radio.net/2018/01/se-radio-episode-315-jeroen-janssens-on-tools-for-data-science/
2018-01-23 / 2018-02-11
- Definition: extracting value from data; two forms:
- Analysis that you make, giving additional insight.
- Integrate into product (eg, recommendations in Netflix).
- What's different about data science from statistics? Statistics is a part of data science.
- Statistics + dealing with data that is messy and getting it into the right form + visualizing
- JJ not a fan of the term "science" in "data science"
- Building a recommendation system is not really "science" because doesn't have hypothesis, etc
- Could be making observations and drawing conclusions, but that is not necessarily what you're doing with data science
- JJ not a fan of "big data" term
- SE 286 about machine learning: http://www.se-radio.net/2017/03/se-radio-episode-286-katie-malone-intro-to-machine-learning/
- Don't set out to learn data science
- Instead, aspire to tackle a certain problem that requires data, and find the skills you need to solve the problem
- Characteristics of programming languages that are good for:
- Analyzing
- Need to be able to play with the data
- Language gets out of your way
- Not Java
- Suitable languages: Python, R
- REPL
- Interesting questions about intellectual property & open source of learning models and results of learning experiments
https://changelog.com/podcast/274
2017-11-18 / 2017-12-10
- Language-independent backgroun jobs
- Sidekiq for other languages
- Invert sidekiq architecture
- Sidekiq talking to Redis
- Redis is 'dumb' has no logic
- Logic is in Sidekiq
- Faktory is like Redis
- Instead of exposing data storage features, exposes job processing features
- Announced October 2017
- Already jobs implemented in several languages
- Data store
- Redis allows "modules" but they have to be compiled with it
- Want a single binary
- 'Embedded' RocksDB from Facebook
- Discussion of how to choose your tool (in this case data store)
https://changelog.com/podcast/255
2017-06-30 / 2017-07-08
- Solve some problems with overly verbose REST APIs
- In particular, REST APIs using Hypermedia
- Avoid multiple requests to different endpoints to get a complete object
- Avoid classic N+1 query as a result
- Need new ways to use HTTP for caching, because doesn't work the same way as REST
- Needs more investigation
- Can be used with other than HTTP
- Does that mean it's not RESTful?
- "Mutations" for create/update
- Time 30 - 45 minutes of the show is about how Graphcool does open source
http://www.se-radio.net/2017/06/se-radio-episode-293-yakov-fain-on-angular/
2017-06-06 / 2017-07-08
- Relies on RxJS
- Has some concept of "observable" which is used where promises used to be used in AngularJS
- Can be used on mobile, like React Native?
http://www.se-radio.net/2017/06/se-radio-episode-295-michael-feathers-on-legacy-code/
2017-06-27 / 2017-07-03
- Difficulty testing can indicate architectural problems
- Concept of "seams"; book contains strategies for dealing with them in different languages
- Problem with Agile: doesn't tell us the cost of change:
- The business only cares about features
- Developers are responsible for code & code health
- Business doesn't know how much it costs to get its features