Machine & Deep Learning Compendium

Searchβ¦

Distribution

- 1.

- 2.
- 3.
- 4.
- 5.
- 6.
- 8.β

- 1.
**It is the most common distribution in nature (as distributions go)** - 2.
**An enormous number of statistical relationships become clear and tractable if one assumes the normal.**

**the Normal distribution in statistics is a special world in which the math is straightforward and all the parts fit together in a way that is easy to understand and interpret.****It may not exactly match the real world, but it is close enough that this one simplifying assumption allows you to predict lots of things, and the predictions are often pretty reasonable.****statistically convenient.****represented by basic statistics****average****variance (or standard deviation) - the average of what's left when you take away the average, but to the power of 2.**

- 1.
**Categorical data can be transformed to a histogram i.e., #class / total and then measured for distance between two histogramsβ, e.g., train and production. Using earth mover distance****python****git wrapper to c****, linear programming, so its slow.** - 2.
- 3.
- 4.
**Also check KL DIVERGENCE in the information theory section.** - 5.β
**Bengio****et al, transfer objective for learning to disentangle casual mechanisms - We propose to meta-learn causal structures based on how fast a learner adapts to new distributions arising from sparse distributional changes**

Last modified 9mo ago

Copy link