Topic Tag: neural networks

Topic 7: modeling best practices (E)

It's now time for you to practice on your own! Start by downloading the materials, if you have not already done so. The required files include: The exercise file (lesson3-topic7-modeling-best-practices-practice.ipynb). This notebook contains 13 steps to guide you on the modeling of best practices. The MAGIC dataset (magic04.csv). This dataset contains simulated information about the registration... Read more »

Topic 6: accelerating neural network training

To learn more about strategies that can accelerate neural network training, watch the 6-minute video below. You can follow along with the instructor’s presentation using the slides available here. The additional resources that were mentioned during the presentation are linked below. Take some time to look them over. What's the Difference Between a CPU and a... Read more »

Topic 5: tuning a model with Keras Tuner (CW)

In this 13-minute video, Jon-Cody will walk you through hyperparameter tuning using Keras Tuner. To follow along with him, ensure that you have found the Topic 5 presentation. You’ll also want to open your anaconda-navigator, set your environment to nasa-nn, launch Jupyter Notebook, and open the notebook we've provided.

Topic 4: visualizing model performance using TensorBoard (CW)

In this 9-minute video, Jon-Cody will build a neural network model using TensorFlow, evaluate it, and visualize the results using TensorBoard. To follow along with him, ensure that you have found the Topic 4 presentation. You’ll also want to open your anaconda-navigator, set your environment to nasa-nn, launch Jupyter Notebook, and open the notebook we've... Read more »

Topic 3: assessing and improving the fit

The 12-minute video below will cover the difference between underfitting and overfitting, and then discuss the most common methods for combatting overfitting in neural networks. To follow along with Jon-Cody’s presentation, download the slides available here. If that video raised any questions for you, post them to the course forum. Otherwise, spend some time reviewing the... Read more »

Topic 2: data prep and model performance metrics (CW)

In this 13-minute video, Jon-Cody will walk you through best practices for data preparation as it relates to neural networks and show you how to compute and evaluate metrics. To follow along with him, ensure that you have found the Topic 2 presentation. You’ll also want to open your anaconda-navigator, set your environment to nasa-nn, launch... Read more »

Topic 1: loss functions and model performance

We start Lesson 3 with a 22-minute video on loss functions and model performance metrics. To follow along with Jon-Cody's presentation, download the slides available here. For easy reference, here is the chart Jon-Cody used to explain when to use the various loss functions: Now, don't forget to spend some time reviewing the descriptions of... Read more »

Topic 7: implementing a custom neural network architecture (E)

It's now time for you to practice on your own! Start by downloading the materials, if you have not already done so. The .zip file will contain: The exercise file (lesson2-topic7-improving-a-nn-practice.ipynb). This notebook contains 15 steps to guide you through the process of improving your neural network using the components that we've been discussing. The... Read more »

Topic 6: implementing a custom neural network architecture (CW)

In the first 3-minute video, Jon-Cody will cover some background information before opening his Python environment in the second video for a 25-minute coding walkthrough. He will create models with different numbers of epochs, batch sizes, and learning rates to compare how those parameters impact loss and accuracy. To follow along with him, ensure that... Read more »

Topic 5: batch size and number of epochs

Watch the 7-minute video that follows for an overview of batch size and epochs and why they are used. You can follow along with the instructor's presentation using the slides available here. Topics 6 and 7 will begin to explore how a model changes when these parameters are adjusted. But before you move on to... Read more »