Part 1: Deep Learning Introduction
In this article, we are talking about the Introduction of Deep Learning which related to neural networks.
(1) What is Deep learning?
- Deep learning is a subset of Machine learning in which multi-layered neural networks learn from a vast amount of data.
- It is a high-level version of Machine learning which uses Artificial Neural Networks as trainable algorithms.
(2) What are the applications of the Deep learning concept?
- Face Recognition
- Self Driving Cars.
- Image Classification
- Medical Diagnosis
- Ads, Search social Recommendations
3. What are the tools mainly used for design & train Deep Learning Model?
- Development Environment: Anakonda Navigator(Jupyter Notebook or Google Colab)
- Modules-Tensorflow, Keras, ScikitLearn, OpenCV, Numpy, Matplotlib, Pandas.
- Programing Language — Python
4. What are the main Categories of Deep Learning?
a. Supervised Deep Learning — Learned in the past to new data using labeled examples to predict future events.
b. Unsupervised Deep Learning- Used when the information used to train is neither classified nor labeled. This studies how systems can infer a function to describe a hidden structure from unlabeled data.
c . Reinforcement Deep Learning — Agent learns in an environment to achieve a long-term goal by maximizing short-term rewards.
5. Neural Networks types according to catergory of Deep Learning (Considering the Algorithm)
a. In Supervised Learning, we can mainly identify 4 categories of Neural Networks.
- Feed-Forward Neural Network — Simple Problems
- Convolutional Neural Networks (CNN) — Pattern Recognition, Image recognition, character recognition
- Recurrent Neural Network (RNN) — time-series data analysis, Stock market analysis, chatbot, Voice related data
- Encoder-decoder architectures -hybried, also includes Capsule and residual neural networks
b. In Unsupervised Learning — there are mainly 2 parts
- Autoencoder -Noise removal/filtering
- Generative Adversarial- fake data/fake face generator, graphic design, for creating datasets
c. In Reinforcement learning- there is only one that can categorize for this.
- Network for actions, values, policies, and models -Deep Q
This article is mainly about introduction about Deep Leaning, and categories with its related neural networks. The next article is mainly focusing on Supervised learning with Feed-Forward Neural Network (FFNN)which is commonly used in deep learning applications.