pacmap_Rnotebook_example.Rmd

---
title: "PaCMAP Example with Reticulate"
output:
  pdf_document: default
  html_notebook: default
---

This notebook demonstrates the ability to utilize PaCMAP in R with the 
[reticulate package](https://rstudio.github.io/reticulate/).

To utilize PaCMAP, you need to install PaCMAP on a local python environment. 
The following cell assumes that you have installed PaCMAP in a 
[conda](https://docs.conda.io/en/latest/) environment. We strongly recommend you
to use conda to manage your environment.

You can install PaCMAP via executing the following commands in your terminal:
- `conda install pacmap -c conda-forge` through conda-forge
- `pip install pacmap` through PyPI.

## Loading the libraries
First, let us load the required libraries.

```{r}
# Install Reticulate if you haven't done so
# install.packages("reticulate")
reticulate::use_condaenv("your_env_name", conda="your_conda_executable_path", required = TRUE)
pacmap <- reticulate::import("pacmap")
```

Loading PaCMAP can lead to fatal error if your PATH variables have not been
configured properly. In some cases, it could lead to error such as:
`Intel MKL FATAL ERROR: Cannot load libmkl_intel_thread.1.dylib`. Such error
can be resolved by removing MKL references in your local environment with the
following commands:

```{bash}
conda remove mkl
conda install nomkl
conda install pacmap  # install pacmap without MKL references
```

## Loading the data

Then, let us load some sample data. In this example, we use the [mammoth dataset](https://github.com/PAIR-code/understanding-umap/tree/master/raw_data). 
The dataset has been stored as a CSV file. `read.csv()` by default will load the
file into a `list`, which will be interpreted as a Python `dict` when converted 
by reticulate. PaCMAP is unable to read the `dictionary`. Therefore, we convert 
the data into a `matrix`, such that it can be properly converted as a 
`numpy.ndarray` object and read by PaCMAP.

```{r}
data <- read.csv("mammoth_data.csv")
data_vector <- unlist(data)
# Convert the vector into a matrix
data_matrix <- matrix(data_vector, ncol = length(data))
```

## Running Dimensionality Reduction

Once we obtained the data in the `matrix` format, we can perform the
dimensionality reduction easily.

```{r}
# Initialize PaCMAP instance
reducer <- pacmap$PaCMAP()

# Perform dimensionality Reduction
embedding <- reducer$fit_transform(data_matrix)
```

Finally, let us visualize the embedding we got.

```{r}
# Visualize the result
library(ggplot2)
visualizeMatrixScatterplot <- function(matrix, dotSize = 1) {
  # Extract the x and y coordinates from the matrix
  x <- matrix[, 1]
  y <- matrix[, 2]

  # Create a scatterplot with custom dot size
  plot(x, y, pch = 19, col = "blue", cex = dotSize, 
       main = "Scatterplot of 2D Matrix", xlab = "X-axis", ylab = "Y-axis")
}

visualizeMatrixScatterplot(embedding, 0.5)
```