[MINI] Feed Forward Neural Networks
Data Skeptic24 Maalis 2017

[MINI] Feed Forward Neural Networks

Feed Forward Neural Networks

In a feed forward neural network, neurons cannot form a cycle. In this episode, we explore how such a network would be able to represent three common logical operators: OR, AND, and XOR. The XOR operation is the interesting case.

Below are the truth tables that describe each of these functions.

AND Truth Table Input 1 Input 2 Output 0 0 0 0 1 0 1 0 0 1 1 1 OR Truth Table Input 1 Input 2 Output 0 0 0 0 1 1 1 0 1 1 1 1 XOR Truth Table Input 1 Input 2 Output 0 0 0 0 1 1 1 0 1 1 1 0

The AND and OR functions should seem very intuitive. Exclusive or (XOR) if true if and only if exactly single input is 1. Could a neural network learn these mathematical functions?

Let's consider the perceptron described below. First we see the visual representation, then the Activation function , followed by the formula for calculating the output.

Can this perceptron learn the AND function?

Sure. Let and

What about OR?

Yup. Let and

An infinite number of possible solutions exist, I just picked values that hopefully seem intuitive. This is also a good example of why the bias term is important. Without it, the AND function could not be represented.

How about XOR?

No. It is not possible to represent XOR with a single layer. It requires two layers. The image below shows how it could be done with two laters.

In the above example, the weights computed for the middle hidden node capture the essence of why this works. This node activates when recieving two positive inputs, thus contributing a heavy penalty to be summed by the output node. If a single input is 1, this node will not activate.

Universal approximation theorem tells us that any continuous function can be tightly approximated using a neural network with only a single hidden layer and a finite number of neurons. With this in mind, a feed forward neural network should be adaquet for any applications. However, in practice, other network architectures and the allowance of more hidden layers are empirically motivated.

Other types neural networks have less strict structal definitions. The various ways one might relax this constraint generate other classes of neural networks that often have interesting properties. We'll get into some of these in future mini-episodes.

Check out our recent blog post on how we're using Periscope Data cohort charts.

Thanks to Periscope Data for sponsoring this episode. More about them at periscopedata.com/skeptics

Jaksot(588)

Primate Poses

Primate Poses

During this season we have talked with researchers working to utilize machine learning for behavioral observations. In previous episodes, you have heard about the software people like Richard use, but you haven't heard much from scientists modifying and using these tools for specific research cases. PhD student, Richard Vogg, is working with multi-camera set-ups to track lemurs and macaques solving puzzle boxes in the wild. His work is part of a larger movement to automate behavioral analyses of video data. Listen in and learn why this tech is useful and why multi-camera setups are a good idea for more reliably identifying poses and individual animals.

31 Heinä 202432min

Generating 3D Animals with YouDream

Generating 3D Animals with YouDream

Generative AI can struggle to create realistic animals and 2D representations often have mistakes like extra limbs and tails. If 2D wasn't hard enough, there are researchers working on generative 3D models. 3D models present an extra challenge because there is paucity of training datasets.In this episode, PhD students Sandeep and Oindrila walked us through their work on creating 3D animals using 2D data. Join us to learn about their pipelines, quality control, tie in with iNaturalist, and how this tech could streamline FX pipelines.

23 Heinä 20241h

Weird Communication

Weird Communication

Today, we sat down with Dr. Ignacio Escalante Meza to learn about opiliones and treehoppers. Opiliones, known as "daddy long legs" in the US, are understudied arachnids known for their tenacious locomotor behavior, sociality, and chemical communication. Treehoppers communicate through the stems of plants using vibrations. They can signal danger, attract mates, and communicate with their offspring. Join us to learn how researchers turn their vibrations into sound waves and study what they have to say.

15 Heinä 202438min

Reducing the Impact of Ship Noise on Marine Mammals

Reducing the Impact of Ship Noise on Marine Mammals

Human shipping operations have increased significantly in the past few decades. While that means international trade and cheap goods for humans, it also means the ocean has experienced an increase in noise pollution. This has a measurable negative impact on marine mammals and other aquatic life. Could mathematics be the solution? This interview explores how optimization techniques can guide voyage optimization in a way that handles multiple optimization objectives including fuel cost and sound reduction.

1 Heinä 202436min

Analysis of Unstructured Data

Analysis of Unstructured Data

Robbie Moon from the Georgia Tech Scheller College of Business joins us to discuss the analysis of unstructured data and the application of NLP methodologies towards financial data.

28 Kesä 202427min

iNaturalist

iNaturalist

Have you ever participated in citizen science? Do you want to? One of the most popular platforms for crowdsourcing biodiversity data is iNaturalist. In addition to being a great science tool, the iNaturalist app can help you identify the organisms you encounter every day. We talked to Executive Director Scott Laurie about how scientists use iNaturalist. We also got to discuss what makes iNaturalist's AI species recognition so good, and how citizen scientists are constantly providing high-quality training data. Listen in and learn how this fun-to-use tool works, where it's headed, and how you can get involved.

24 Kesä 202437min

Learn to Code

Learn to Code

Do you code or are you interested in learning to code? Join us today and hear from three individuals that are at very different stages of their coding journeys. Becky Hansis-O'Neill (also our co-host this season) shares her experiences as a newbie who wants to learn more. Dr. Malia Gehan, a self-taught developer interested in studying plant phenotypes, explains why and how she and her colleagues learned to code and developed PlantCV. Finally, Dr. John Wilmes discusses his work as a professional mathematician and Machine Learning Research Engineer. Whether you are thinking about learning to code or an expert, we're sure you will see a bit of yourself in this episode.

18 Kesä 202449min

Animal Computer Interaction

Animal Computer Interaction

You've heard of Human Computer Interaction (HCI), now get ready for Animal Computer Interaction (ACI). Ilyena has made a career developing computer interfaces for non-human animals. She has worked with dogs, parrots, primates, and even giraffes. This is challenging because animals have a wide range of abilities and preferences. Parrots, for example, use their tongues to make selections on touchscreens. Listen in on our conversation and learn about interface development and testing with animals and how technology may improve animal welfare.

10 Kesä 202442min

Suosittua kategoriassa Tiede

rss-mita-tulisi-tietaa
utelias-mieli
tiedekulma-podcast
hippokrateen-vastaanotolla
rss-poliisin-mieli
docemilia
sotataidon-ytimessa
filocast-filosofian-perusteet
rss-lihavuudesta-podcast
rss-duodecim-lehti
menologeja-tutkimusmatka-vaihdevuosiin
rss-ammamafia
rss-tiedetta-vai-tarinaa
rss-ilmasto-kriisissa
vinkista-vihia
radio-antro
rss-ranskaa-raakana
rss-jyvaskylan-yliopisto
rss-pandapodi