Create page with info about trainers

Currently, there is not a unified page to go to to see a summary of what each trainer is good at, and its features/limitations.

Here's a start which I made with @glebuk a while back:

The cells without information (empty cells) are simply not filled in. Specifically, it's a missing answer.

In Excel form:
[ML.NET Trainer Cheatsheet.xlsx](https://github.com/dotnet/machinelearning/files/6917619/ML.NET.Trainer.Cheatsheet.xlsx)

# Table in Markdown form

| | | Tree Ensemble | Linear | META | Baseline | Other | |
| -------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------- | ---------- | --------- | -------- | ------- | ------- |
| | | GAMs | FastForest | Fast Tree | LightGBM | SDCA | Sym SGD | LbFGS | Averaged Perceptron | OlsTrainer | OGD | OVA | Pairwise Coupling | Prior Trainer | KMeansTrainer | PCA | Linear SVM | Field Aware FM | Naiive Bayes | |
| Tasks | Binary Classification | ★ | ★ | ★★ | ★★ | ★ | ★★ | ★ | ★ | | | | | ★ | | | ★ | ★ | | |
| Multi-class Classification | ★ \[3\] | ★ \[3\] | ★ \[3\] | ★★ | ★★ | ★ \[3\] | ★ | ★ \[3\] | | | ★ | ★ | ★ \[3\] | | | ★ \[3\] | | ★ | |
| Regression | ★ | ★ | ★ | ★ | ★ | | ★ | | ★ | ★ | | | | | | | | | |
| Clustering | | | | | | | | | | | | | | ★ | | | | | |
| Anomaly Detection | | | | | | | | | | | | | | | ★ | | | | |
| Ranking | | | ★★ | ★ | | | | | | | | | | | | | | | |
| Good for Features | Ngram Featurized Text | ❌ | ★ | ★★ | ★★ | ★ | ★ | ★ | ★★★ | | | | | ❌ | | | ★ | ★★ | ❌ | |
| Word Embedding | ❌ | ★ | ★★ | ★★ | ★★★ | ★★ | | ★★★ | | | | | ❌ | | | | | ❌ | |
| Categorical | | ★ | ★★ | ★★ | ★★ | ★★ | ★★ | ★★ | | | | | ❌ | | | | ★★★ | ❌ | |
| Numeric | | ★ | ★★ | ★★ | | | | | | | | | ❌ | ★ | | | | ❌ | |
| Featurized Image | ❌ | ★ | ★★★ | ★★ | ★★ | ★ | | | | | | | ❌ | | | | | ❌ | |
| Accuracy | Good first try | ❌ | ❌ | ★★ | ★★ | ★ | ★ | ❌ | ★★ | | | ★ | | | | | ❌ | | ❌ | |
| Average Accuracy | | ★★ | ★★★ | ★★★½ | ★★ | ★★ | | | | | | | | | | | ★★ | ½ | |
| Is Tunable | | | ★★★ | ★★ | ★ | | ★ | | | | ❌ | ❌ | ❌ | | | | | ❌ | |
| Speed | Speed | ½ | ★ | ★★ | ★★½ | ★★★ | ★★★★ | ★★ | ★★★★ | | | | | ★★★★ | ★★ | ½ | ½ | ★★★ | ★★★★ | |
| Streaming (data > RAM) | ❌ | ❌ | ❌ | ❌ | ★ | ★ | | ★ | | | ★ \[2\] | ★ \[2\] | ★ | ★ | | ★ | ★ | ★ \[4\] | |
| RAM (bytes) per Row | row | row | row | row | 8 | ∅ | row | ∅ | row | row | | | ∅ | feat\*k\*8 | row | row | ? | ∅ | |
| ~Scalability for #features | 10^5 | 10^5 | 10^5 | 10^5 | 10^9 | 10^9 | 10^9 | 10^9 | | | | |  ∞ | | 10^4 | 10^5 | ? | | |
| Benefits from Caching | ❌ | ❌ | ❌ | ❌ | ★ | ★? | | ★ | | | | | | ★ | | ★ | | ❌ | |
| Model | Explainable Model | ★★★ | ★ | ★ | ★ | ★★ | ★★ | ★★ | ★★ | ★★ | ★★ | | | | ? | | | | ★ | |
| Exportable to ONNX | | ★ | ★ | | ★ | ★ | ★ | ? | | | | | | ★ | | | | | |
| Can be turned into Code | ★ | ★ | ★ | ★ | ★ | ★ | ★ | ? | | | | | ★ | ? | | | | | |
| Requires Normalization | ❌ | ❌ | ❌ | ❌ | ★ | ★ | ★ | ★ | | | | | | ★ \[1\] | ★? | ★ | ? | ★ \[5\] | |
| Returns Probability | ❌ | ❌ | ❌ | ❌ | ★ | | ★ | ❌ | | | | | | ❌ | ❌ | | | ★ | |
| Online Updatable | ❌ | ❌ | ❌ | ❌ | ★ | ★ | ★ | ★ | | | | | | ? | | ★ | ★ | | |
| Model Size (bytes/class) | ~ | ~ | ~ | ~ | feat\*8 | feat\*8 | feat\*8 | feat\*8 | | | feat\*8 | feat\*8 | 0 | feat\*8\*k | | feat\*8 | | feat\*8 | |
| | | | | | | | | | | | |

Notes:
\[1\] k-means needs normalization, but you can also up weight certain features by giving them a larger scale
\[2\] ova & pkpd are streamable if the underlying trainers are
\[3\] most binary trainers handle multi-class classification tasks when used with OVA or PKPD
\[4\] naive bayes stops reading at 2B rows (is fixed?)
\[5\] naive bayes does better on numeric data with MeanVar w/ fixZero=false, or whitening (both move the mean to zero; though will densify ngrams/categorical which is very slow)