NMF.Rmd

---
title: "NMF"
author: "liuc"
date: "11/10/2021"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```

## R Markdown

NMF的思想：V=WH（W权重矩阵、H特征矩阵、V原矩阵），通过计算从原矩阵提取权重和特征两个不同的矩阵出来。属于一个无监督学习的算法，其中限制条件就是W和H中的所有元素都要大于0.
W又称为基矩阵、H称为系数矩阵。
```{r}
library(NMF)

# introduction to NMF package
# 输入矩阵为常见的gene counts矩阵
# For example, if min(transposed_data) outputs -3.1, I add 3.100001 to the matrix

nmfAlgorithm() # 查看NMF包内的算法
nmfAlgorithm('brunet')

nmfSeed() # 查看seed的用法
# number value is used to set the state of the RNG, and the initialization is performed by the built-in seeding method ’random’
# Note that if the seeding method is deterministic there is no need to perform multiple run anymore
```


```{r}
data(esGolub)

# 为避免过拟合，此处数据可randomized
V.random <- randomize(esGolub)

estim.r <- nmf(esGolub, 2:6, 
               nrun=30,
               seed=123456
               )
plot(estim.r)
# plot(2:6,estim.r$measures$cophenetic, 
#      type="b",col="purple")

# 判断最佳rank值的准则就是，cophenetic值随K变化的最大变动的前点

consensusmap(estim.r,
             annCol=esGolub, labCol=NA, labRow=NA
             )
```

```{r}
 nmfEstimateRank(x, range,
    method = nmf.getOption("default.algorithm"), nrun = 30,
    model = NULL, ..., verbose = FALSE, stop = FALSE)
```

```{r}
res1 <- nmf(x = esGolub, 
            rank = 4, 
            nrun=100,
            method = "brunet",
            seed=123456)

w <- basis(res1)
h <- coef(res1)
V.hat <- fitted(res1)

fit(res1) # model info
summary(res1) #  quality measures 
summary(res1, class=esGolub$Cell)


dim(res1[1:10, 1:10])
```
特征提取
```{r}
# only compute the scores
s <- featureScore(res1)
summary(s)


# compute the scores and characterize each metagene
s <- extractFeatures(res1)
str(s)


```

comparing algorithms
```{r}
#  To enable a fair comparison, a deterministic seeding method should also be used. Here we fix the random seed to 123456.

res.multi.method <- nmf(esGolub, 3, list('brunet', 'lee', 'ns'), 
                        seed=123456, 
                        .options='t')
compare(res.multi.method)
```

# 针对W H矩阵的热图绘制
```{r}
# basis components
basismap(res1, subsetRow=TRUE)
# mixture coefficients
coefmap(res1)
```