Question 1

What is an example of deep learning?

Accepted Answer

Some examples of deep learning include virtual assistants, self-driving cars and facial recognition software.

Deep learning is a subfield of machine learning that utilizes algorithms to abstractly mimic the structure and general functionality of the human brain, building what are known as artificial neural networks to transmit information in non-linear ways. Deep learning relies on the input of vast quantities of data in order to begin building knowledge within a system and training it on proper decision making while it scales.

Though artificial neural networks require heavy training to provide value once implemented, the technology offers bounteous potential with little human intervention required, especially when it comes to analyzing the ever-growing quantity of raw, unlabeled data in the world. This is especially useful when it comes to analyzing data from non-tabular based sources, such as video and image content. Some deep learning examples include:

Fraud detection software, such as programs that monitor credit card purchases to detect unauthorized usage
Self-driving cars
Facial recognition software
Natural language processing that can translate text or power chatbots
Recommendation engines such as Netflix, which uncover patterns in a user’s viewing behavior and recommend new shows to watch

Question 2

What is deep learning vs machine learning?

Accepted Answer

Deep learning is a subcategory of both machine learning and artificial intelligence that is unique in its ability to handle analog inputs and outputs.

Where deep learning excels over machine learning comes largely in its ability to understand data that does not come in tabular form, such as pixel data, text documents and audio files. One type of deep learning that is concerned entirely with images is known as a convolutional neural network, which applies weight-based filters to every element in an image to allow the computer to understand and react to the picture. Another form of deep learning is a recurrent neural network, which incorporates memory in order to keep past decision points in mind when reviewing recurring data.

Other ways in which deep learning differs from machine learning is in deep learning’s lack of necessary ongoing human intervention, need for more intense hardware such as powerful graphical processing units, time required to set up, efficiency in producing instantaneous results, use of unstructured data and use in complex autonomous programs.

Question 3

Is LSTM deep learning?

Accepted Answer

Long Short-Term Memory (LSTM) is a category of neural network that allows computers to learn order dependence in sequence prediction.

Long Short-Term Memory (LTSM) is a type of recurrent neural network deep learning that trains computers to be able to learn order dependence in problems that require sequence prediction.

LTSMs work by creating repeating modules with interacting layers to add temporal dependencies into the equation and allow every element of an image or other data source to be analyzed. This is particularly useful when determining similarities and differences between large batches of data in order to produce a more nuanced output for each individual data point.