What is Neural Network?
Neural Networks are one of the dominant machine learning technologies. If you are thinking that this is something like human brains, you’re almost there. Neural networks are at the heart of Deep learning which reflects or mimics the behavior of a human brain with N number of neurons containing input with weights, weighted sum and bias, activation function, and output connected.
When a bunch of such perceptron are connected it makes an Artificial Neural Network (ANN)
Before going further into neural networks, let’s understand
What is Deep Learning?
Deep learning is a subset of Machine Learning in Artificial Intelligence where Artificial neural networks are used to learn vast data and makes multilayer neural networks feasible by attempting to simulate the behavior of human brain by connecting multiple perceptron as layers.
How does this Work?
Deep neural network consists of the multiple layers with nodes interconnected where the Input Layer are the visible layers, where the data is ingested, Hidden layers, where the feature extraction, etc. is done and finally the output layer, where the classification or predictions are done.
The progression of computations through this network, building on the previous layers as it moves forward optimizing the prediction is called forward propagation. Similarly, backpropagation is a process used to calculate the error in the prediction, function bias, adjusts the weights, moving backward through the layers during the model training, usually done with algorithms like gradient descent. The forward and backpropagation together gradually improve the model accuracy, allowing neural network to correct the errors and make predictions.
This has a great impact on real-world applications like self-driving cars, voice-enabled AI assistance, fraud detections, improving automation, etc.
Deep Learning Vs Machine Learning
As I mentioned above, Deep learning is a subset of Machine Learning in AI.
Classical machine learning algorithm works on structured, labeled data where there are specific features and the target variables. It also works on unstructured data with some pre-processing to format the data.
On the other hand, deep learning algorithms work well on unstructured and massive data such as voice, text, images, etc. Deep learning reduces the human dependency / manual work on data pre-processing and automates feature extraction.
Neural Network architectures in deep learning
Artificial Neural Network (ANN): As we already saw, it’s a traditional neural network and a functional unit of deep learning, it is a feed-forwarding neural network as the inputs gets processed only in the forward direction i.e., only forward propagation. It can be used in Tabular data, text data but it’s recommended for Image data as it loses spatial features, the images should be converted to a 1-dimensional vector before training the model.
Convolution Neural network (CNN): This is an eye-catcher, especially when it comes to Image data as it captures spatial features in the image and identifies the object, location of the target object. It automatically learns the filters to extract the features that are relevant to the input data without mentioning explicitly. This is widely used for applications like Image classification, image segmentation, object detection, etc.
Check out this interesting application of CNN in astronomy.
Stay Tuned for an elaborate explanation on CNN layers, ImageNet, AlexNet, CNN working with an example…
Recurrent Neural Network (RNN): This is a parameter sharing concept like CNN with recurrent connections. In simple words, it is an ANN model with looped neurons in the hidden layers. You would have heard of LSTM – it is a RNN model, which solves tasks involving sequential data like Time series, Text, and audio which are not solvable by the above algorithms.
We will discuss LSTM and its working with real-time use cases, in our future blogs.
Stay Tuned for more freaky news and facts. Feel free to share your thoughts in the comment section and do check our other posts on cyber and AI here.
Follow us on Instagram for the latest updates in Cybersecurity and AI.
You can find me on LinkedIn
Bình luận