Deep Learning with Neural Networks- Part 1

Udara Vimukthi
3 min readJun 9, 2021


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.



Udara Vimukthi

Final Year Undergraduate, University of Moratuwa.