When Sat Nov 27 4pm — 6pm
Where Classroom - Interactive
1825 Monetary Ln #104 Carrollton, TX 75006
What Advantages of using Hidden Layers, Multilayer Perceptrons, Forward and Backward Propogation, Model Selection, Initialization, Weight Decay, Dropout
Host Kevin Thompson


  • Cost: Free

Cancellations for this event must be made before November 27, 2021 — 4:00pm.

Registration for the event is closed.

About this Event

In Deep Learning Prerequisites, we went over the mathematical and computational foundations of Deep Learning.

In Deep Learning Part 2: Information Theory and Linear Models, we went over the theoretical foundations of Deep Learning.

In Deep Learning Part 3: Multilayer Perceptrons Done Right, what we learned in the first two sections payoff spectacularly when we are introduced to the most fundamental Deep Neural Network (DNN) architecture: the multilayer perceptron. We will discuss core properties of the multilayer perceptron, its implementations, how to select the best model configuration, and some of the key techniques that are used to train DNNs in industry today.

Recommended (but not required) prior reading:

1) Chapter 4 of Dive Into Deep Learning (excluding 4.9 and 4.10) - https://d2l.ai/chapter_multilayer-perceptrons/index.html

2) I also strongly recommend reading the "Evaluating and Comparing Estimators" section in the appendix of the free online textbook https://d2l.ai/chapter_appendix-mathematics-for-deep-learning/statistics.html#evaluating-and-comparing-estimators

Special Considerations and Warnings

As in the previous 2 parts, I highly recommend attending the lecture with a laptop that has anaconda installed. You can find the installer at https://www.anaconda.com/products/individual

Also like in previous parts, I still strongly recommend that you try your best to read the recommended (but not required) prior reading material mentioned in the description.