jrafolsr / tensorflow-pokemon-course

• 👋 Hi, I’m @jrafolsr
• 👀 I’m interested in experimental physics, specially organic electronics, but I also enjoy creating and coding new scientific instruments using Python.
• 🌱 I’m currently learning how to create python libraries and modules that could eventually be useful to other experimental physicists.
• 💞️ I’m looking to collaborate on any project that involves controlling lab instruments with python (e.g. pyvisa, pyhton-seabreeze or similar) and potentially even creating some user-friendly GUIs (currently using dash, from plotly).

tensorflow-pokemon-course's Issues

Compile and Evaluate Model

Once we have decided on the specifics of our model, we need to do two processes: Compile the model and fit the data to the model.

We can compile the model like so:

model.compile(optimizer='sgd', loss='sparse_categorical_crossentropy', metrics=['accuracy'])

Here we're just feeding three parameters to model.compile. We pick an optimizer, which determines how the model is updated as it gains information, a loss function, which measures how accurate the model is as it trains, and metrics, which specifies which information it provides so we can analyze the model.

The optimizer we're using is the Stochastic Gradient Descent (SGD) optimization algorithm, but there are others available. For our loss we're using sparse_categorical_crossentropy. If our values were one-hot encoded, we would want to use "categorial_crossentropy" instead.

Then we have the model fit our training data:

model.fit(train_data, train_labels, epochs=400)

The three parameters model.fit needs are our training data, our training labels, and the number of epochs. One epoch is when the model has iterated over every sample once. Essentially the number of epochs is equal to the number of times we want to cycle through the data. We'll start with just 1 epoch, and then show that increasing the epoch improves the results.

True or False: We fit our model using test_data and test_labels.

Install and setup environment

Welcome to the world of machine learning with TensorFlow! Working with TensorFlow can seem intimidating at first, but this tutorial will start with the basics to ensure you have a strong foundation with the package. This tutorial will be focusing on classifying and predicting Pokémon, but the elements discussed within it can certainly be helpful when using TensorFlow for other ideas, as well. Without further ado, let's begin!

First, let's download TensorFlow through pip. While you can install the version of TensorFlow that uses your GPU, we'll be using the CPU-driven TensorFlow. Type this into your terminal:

pip install tensorflow

Now that it's installed, we can truly begin. Let's import Tensorflow, and a few other packages we'll need. All of this course involve using the command line interface. Enter these commands to import and the necessary packages:

import tensorflow as tf
from tensorflow import keras
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn import preprocessing

Leave a comment with your favorite Pokémon (such as Pikachu) to continue.

Split and Normalize Data

Now that we have our data in a useable form, we need to split it. We want to have a set of data that we'll use to train our model, and we'll use another set of data to test our model after we've trained it. In general, the data is randomly split with about 70% being used for training and 30% used for testing. For easier visualization, we'll be splitting the data by Pokémon generation. The first generation of Pokémon (from Pokémon Red, Blue, and Yellow) will be our testing data while the rest will be our training data:

def train_test_splitter(DataFrame, column):
    df_train = DataFrame.loc[df[column] != 1]
    df_test = DataFrame.loc[df[column] == 1]

    df_train = df_train.drop(column, axis=1)
    df_test = df_test.drop(column, axis=1)

    return(df_train, df_test)

df_train, df_test = train_test_splitter(df, 'Generation')

This function takes any Pokémon whose "Generation" label is equal to 1 and putting it into the test dataset, and putting everyone else in the training dataset. It then drops the Generation category from the dataset.

Now that we have our two sets of data, we'll need to separate the labels (the 'islegendary' category) from the rest of the data. Remember, this is the answer key to the test the algorithms are trying to solve, and it does no good to have them learn with the answer-key in (metaphorical) hand:

def label_delineator(df_train, df_test, label):
    
    train_data = df_train.drop(label, axis=1).values
    train_labels = df_train[label].values
    test_data = df_test.drop(label,axis=1).values
    test_labels = df_test[label].values
    return(train_data, train_labels, test_data, test_labels)

This function extracts the data from the DataFrame and puts it into arrays that TensorFlow can understand with.values. We then have the four groups of data:

train_data, train_labels, test_data, test_labels = label_delineator(df_train, df_test, 'isLegendary')

Comment with the generation number we used in the test dataset.