Deep Learning

What is Deep Learning?

Deep Learning is a subgroup of Machine Learning which is a subgroup of Artificial Intelligence.

What is Artificial Intelligence?

Artificial Intelligence is the ability of machines to emulate the intelligent behavior of humans.

What is Machine Learning?

Machine Learning is a subset of Artificial Intelligence that gives the system the ability to learn from a set of examples without being program explicitly.

Limitations of Machine Learning:

  1. Machine Learning models cannot deal with high dimensional data
  2. Since the machine learning models cannot automatically select the features it becomes difficult to solve complex problems like fingerprint recognition or image recognition.
  3. Deep Learning overcomes the above limitations of Machine Learning. Deep Learning implements with the help of Neural Networks which are motivate by biological neurons.

Biological Neurons

In the biological neurons, dendrites are use to provide inputs to the neuron. Inside a cell body, there’s the nucleus which performs some function. Then the output of this function will travel through the axon and go to the axon terminal and then the neuron produces this output to the next neuron. The study says that two neurons are never connect up. There’s a gap between them; this gap is the synapse.

In the Artificial neural network, there are multiple inputs just like a biological neuron. These inputs multiplies with the weights & provided to the processing element which corresponds to the cell body in the biological neuron. Then summation will happen which will generate transfer function F(S). Then this output is provide to the Activation function. The activation function provides the threshold.

So if your output is above the threshold then only it will fire it otherwise it won’t. There are various activation functions like Step activation function, Sigmoid activation function, etc. Then this actual output compares with the desired output. If the actual output doesn’t match with the desired output then we will find the difference between actual output and desired output and based on this difference we will again update the weights. This process keeps on repeating until we get the desired output. The process of updating weights refer backpropagation.

Deep Learning | artificial neural networks

Deep Learning implements with the help of deep networks and deep networks are the neural networks with multiple hidden layers. In the following figure, there is one input layer, two hidden layers & one output layer. thus, Nodes of each layer interconnects to each other.

There can be hundreds of hidden layers in the Deep Learning model but when we talk about machine learning that was not the case.

Applications:

1) Health care industry:

In the healthcare industry deep learning can also be used for drug prediction & identification of cancer cells.

2) Cell phone /internet industry:

In the cell phone /internet industry we come across various applications for speech recognition & video/image classification. For example, Google Voice, Microsoft Skype, Apple Siri, etc.

3) News/entertainment/media:

In the news/entertainment/media industry it can be used for video captioning, recommendation systems such as amazon & Netflix, real-time translation.

4) Automobile Industry:

In the automobile industry, it can be used in self-driving cars for tracking the path, sign prediction & passenger identification.

5) Security:

Deep learning can be used in various security devices for video surveillance & face recognition.

Machine Learning vs. Deep Learning:

 Machine LearningDeep Learning
Feature selectionHere the user has to find out the best features that are useful for further processing.Here the model is capable of finding out the best future set.
Data sizePerforms well on small or medium size datasets.Performs well on a large dataset.
Hardware requirementIt can work on a low-end system.Deep Learning requires high-end systems having GPU as their parallel architecture helps to reduce the time required for training from large amounts of data or complex matrix computations.
Time for executionIt ranges from minutes to a few hoursIt can be as large as weeks due to the weight calculation process in neural networks.

Conclusion:

In this article, we have learned what is deep learning, artificial intelligence & machine learning? what are the limitations of machine learning and how it can be used to overcome those limitations, its applications, and comparative analysis of deep learning & machine learning?

Written By: Priyanka Shahane

Reviewed By: Rushikesh Lavate

If you are Interested In Machine Learning You Can Check Machine Learning Internship Program
Also Check Other Technical And Non Technical Internship Programs

Leave a Comment

Your email address will not be published. Required fields are marked *