Development environments are personal. Most developers will prefer one code editor or toolset over another. While we recommend the use of interactive tools such as Jupyter via gophernotes, the only prerequisite to running the code examples in this book is a working installation of Go 1.10 or higher. That is, the go command should be available and the GOPATH environment variable should be set up correctly.

To install Go, download a binary release for your system from https://golang.org/dl/. Then, refer to the one of the following subsections that matches your operating system^[2].

Project Jupyter is a not-for-profit organization that was created to develop language-agnostic interactive computing for data science^[3]. The result is a mature, well-supported environment to explore, visualize, and process data that can significantly accelerate development by providing immediate feedback and integrations with plotting libraries such as gonum/plot.

While its first iteration, called iPython, only supported Python-based handlers (called kernels) at first, the latest version of Jupyter has over 50 kernels that support dozens of languages, including three kernels for the Go language^[4]. GitHub has support for rendering Jupyter files (called notebooks)^[5], and there are various specialized hubs for sharing notebooks online, including Google Research Colabs^[6], Jupyter's community hub called NBViewer^[7], and its enterprise...

In our first code example, we will use the multi-domain sentiment dataset (version 2.0)^[11]. This dataset contains Amazon reviews from four different product categories. We will download it, preprocess it, and load it into Gota, a data wrangling library, to find the most common phrases in positive and negative reviews that do not co-occur in both. This is a basic example that involves no ML algorithms, but will serve as a hands-on introduction to Go, gophernotes, and Gota.

In the previous section, we introduced gophernotes and Gota. In this section, we will explore a dataset containing 500 samples of gender, height, and BMI index. We will do this using the gonum/plot library. This library, which was originally a fork of the 2012 Plotinum library^[15], contains several packages that make creating data visualizations in Go much easier^[16]:

• The plot package contains a layout and formatting interface.
• The plotter package abstracts the layout and formatting for common plot types, such as bar charts, scatter plots, and so on.
• The plotutil package contains utility funcs for common plot types.
• The vg package exposes an API for vector graphics and is particularly useful when exporting plots to other software. We will not be covering this package.

The quality and speed of the ML algorithm training process depends on the quality of the input data. While many algorithms are robust to irrelevant columns and data that is not normalized, some are not. For example, many models requires data inputs to be normalized to lie between 0 and 1. In this section, we will look at some quick and easy ways to preprocess data with Gota. For these examples, we will be using a dataset containing 1,035 records of the height (inch) and weight (lbs) of major league baseball players^[17]. The dataset, as described on the UCLA website, consists of the following features:

• Name: Player name
• Team: The baseball team that the player was a member of
• Position: The player's position
• Height (inches): Player height
• Weight (pounds): Player weight in pounds
• Age: Player age at the time of recording

For the...

In this chapter, we covered how to set up a development environment for Go that is optimized for ML applications. We explained how to install an interactive environment, Jupyter, to accelerate data exploration and visualization using libraries such as Gota and gonum/plot.

We also introduced some basic data processing steps, such as filtering outliers, removing unnecessary columns, and normalization. Finally, we covered sampling. This chapter took the first few steps in the ML life cycle: data acquisition, exploration, and preparation. Now that you have read this chapter, you have learned how to load data into a Gota dataframe, how to use the dataframe and series packages to process and prepare the data into a format that is required by your chosen algorithm, and how to visualize it with gonum's plot package. You have also learned about different ways of normalizing...

1. Software Development Waste. Todd Sedano and Paul Ralph. ICSE '17 Proceedings of the 39th International Conference on Software Engineering. Pages 130-140.
2. See the official Go installation instructions at https://golang.org/doc/install. Retrieved February 19th, 2019.
3. https://jupyter.org/about. Retrieved February 19th, 2019.
4. https://github.com/jupyter/jupyter/wiki/Jupyter-kernels. Retrieved February 19th, 2019.
5. For further instructions, see https://help.github.com/articles/working-with-jupyter-notebook-files-on-github/. Retrieved February 19th, 2019.
6. https://colab.research.google.com. Retrieved February 19th, 2019.
7. https://nbviewer.jupyter.org/. Retrieved February 19th, 2019.
8. https://jupyter.org/hub. Retrieved February 19th, 2019.
9. https://github.com/jupyter/nbconvert. Retrieved February 19th, 2019.
10. For Docker installation instructions, see https://docs.docker...