update readme

Author: dselivanov

DESCRIPTION

@@ -5,7 +5,7 @@ Version: 0.2.0
 Author: Dmitriy Selivanov
 Maintainer: Dmitriy Selivanov <selivanov.dmitriy@gmail.com>
 Description: Implements many (sparse) matrix factorizations algorithms. 
-    Focus is applications for recommender systems. 
+    Focus is on applications for recommender systems. 
     Following algorithms are implemented at the moment: 
     1) Weighted Regularazied Matrix Factorization with Alternationf Least Squares (ALS) 
       for implicit feedback (inculding approximate Conjugate Gradient solver).

README.md

@@ -1,23 +1,42 @@
 # reco
 
-`reco` is an R package which implements several algrithms for matrix factorization targeting recommender systems. 
+`reco` is an R package which implements many algorithms for **sparse matrix factorizations**. Focus is on applications for **recommender systems**.
 
-1. Weighted Regularized Matrix Factorization (WRMF) from [Collaborative Filtering for Implicit Feedback Datasets](http://yifanhu.net/PUB/cf.pdf) (by Yifan Hu, Yehuda Koren, Chris Volinsky). One of the most efficient (benchmarks below) solvers.
-1. Linear-Flow from [Practical Linear Models for Large-Scale One-Class Collaborative Filtering](http://www.bkveton.com/docs/ijcai2016.pdf). This algorithm is similar to [SLIM](http://glaros.dtc.umn.edu/gkhome/node/774) but looks for factorized low-rank item-item similarity matrix.
-1. Regularized Matrix Factorization (MF) - classic approch for "rating" prediction.
+## Algorithms
 
-Package is **quite fast**:
+1. Vanilla **Maximum Margin Matrix Factorization** - classic approch for "rating" prediction. See `WRMF` class and constructor option `feedback = "explicit"`. Original paper which indroduced MMMF could be found [here](http://ttic.uchicago.edu/~nati/Publications/MMMFnips04.pdf).
+    * <img src="docs/img/MMMF.png" width="400">
+1. **Weighted Regularized Matrix Factorization (WRMF)** from [Collaborative Filtering for Implicit Feedback Datasets](http://yifanhu.net/PUB/cf.pdf). See `WRMF` class and constructor option `feedback = "implicit"`. 
+We provide 2 solvers:
+    1. Exact based of Cholesky Factorization
+    1. Approximated based on fixed number of steps of **Conjugate Gradient**.
+See details in [Applications of the Conjugate Gradient Method for Implicit Feedback Collaborative Filtering](https://pdfs.semanticscholar.org/bfdf/7af6cf7fd7bb5e6b6db5bbd91be11597eaf0.pdf) and [Faster Implicit Matrix Factorization](www.benfrederickson.com/fast-implicit-matrix-factorization/).
+    * <img src="docs/img/WRMF.png" width="400">
+1. **Linear-Flow** from [Practical Linear Models for Large-Scale One-Class Collaborative Filtering](http://www.bkveton.com/docs/ijcai2016.pdf). Algorithm looks for factorized low-rank item-item similarity matrix (in some sense it is similar to [SLIM](http://glaros.dtc.umn.edu/gkhome/node/774))
+    * <img src="docs/img/LinearFlow.png" width="300">
+1. **Soft-SVD** via fast Alternating Least Squares as described in [Matrix Completion and Low-Rank SVD via Fast Alternating Least Squares](https://arxiv.org/pdf/1410.2596.pdf).
+    * <img src="docs/img/soft-svd.png" width="600">
+1. **Soft-Impute** via fast Alternating Least Squares as described in [Matrix Completion and Low-Rank SVD via Fast Alternating Least Squares](https://arxiv.org/pdf/1410.2596.pdf).
+    * <img src="docs/img/soft-impute.png" width="400">
+    * with a solution in SVD form <img src="docs/img/soft-impute-svd-form.png" width="150">
 
-* Built on top of `RcppArmadillo`
-* extensively use **BLAS** and parallelized with **OpenMP**
-* implements **Conjugate Gradient solver** as dicribed in [Applications of the Conjugate Gradient Method for Implicit
-Feedback Collaborative Filtering](https://pdfs.semanticscholar.org/bfdf/7af6cf7fd7bb5e6b6db5bbd91be11597eaf0.pdf) and [Faster Implicit Matrix Factorization](www.benfrederickson.com/fast-implicit-matrix-factorization/)
-* Top-k items inference is `O(n*log(k))` and use **BLAS** + **OpenMP**
 
-![benchmark](https://github.com/dselivanov/bench-wals/raw/master/img/wals-bench-cg.png)
+## Efficiency
 
-# Tutorials
+Package is reasonably fast and scales nicely to datasets with millions of rows and millions of columns:
 
+* built on top of `RcppArmadillo`
+* extensively uses **BLAS** and parallelized with **OpenMP**
+
+Here is example of `reco::WRMF` on [lastfm360k](https://www.upf.edu/web/mtg/lastfm360k) dataset in comparison with other good implementations:
+
+<img src="https://github.com/dselivanov/bench-wals/raw/master/img/wals-bench-cg.png" width="600">
+
+# Materials
+
+**Note that syntax could be not up to date since package is under active development**
+
+1. [Slides from DataFest Tbilisi(2017-11-16)](https://www.slideshare.net/DmitriySelivanov/matrix-factorizations-for-recommender-systems)
 1. [Introduction to matrix factorization with Weighted-ALS algorithm](http://dsnotes.com/post/2017-05-28-matrix-factorization-for-recommender-systems/) - collaborative filtering for implicit feedback datasets.
 1. [Music recommendations using LastFM-360K dataset](http://dsnotes.com/post/2017-06-28-matrix-factorization-for-recommender-systems-part-2/)
     * evaluation metrics for ranking
@@ -32,7 +51,7 @@ Feedback Collaborative Filtering](https://pdfs.semanticscholar.org/bfdf/7af6cf7f
 
 We follow [mlapi](https://github.com/dselivanov/mlapi) conventions.
 
-# Notes on multithreading and BLAS
+### Notes on multithreading and BLAS
 
 **VERY IMPORTANT** if you use multithreaded BLAS (you generally should) such as OpenBLAS, Intel MKL, Apple Accelerate, I **highly recommend disable its internal multithreading ability**. This leads to **substantial speedups** for this package (can be easily 10x and more). Matrix factorization is already parallelized in package with OpenMP. This can be done by setting corresponding environment variables **before starting `R`**: