Advanced Certificate Programme in AI, ML and DL ( formerly known as Certificate Programme in Machine Learning and Deep Learning ) (AIMLDL Batch 8)

Foundations of Python Programming

Functional Programming in Python

Data Structures, Loops, and Control Structures

Object-oriented programming

Numerical Computations and Linear Algebra using NumPy

Data Pre-processing using Pandas

Data Visualisation using Matplotlib

Introduction to Scikit-learn

Linear Algebra: Vectors, Matrices, Norms, Subspaces, Projections, SVD, EVD, Derivatives of Matrices, Vector Derivative Identities, Least Squares

Optimisation: Constrained and Unconstrained Optimisation, Maxima and Minima, Convex and Non-Convex, Gradient and Hessian, Positive Definite and Semi-Definite, Second Derivative Test, Steepest Descent, Adam, AdaGrad, RMSProp, and KKT

Probability Theory: Discrete and Continuous Random Variables, Conditional Probability, Joint Probability Distribution, Multivariate, MAP Criterion, and ML Criterion

Gaining an understanding of the mathematical fundamentals crucial for machine and deep learning success, like linear algebra, probability theory, and optimisation methods. In linear algebra, one masters essential operations involving vectors and matrices and the understanding of eigenvalues and eigenvectors. Probability theory will provide concepts on probability distributions and Bayes' theorem, which is crucial to understanding the probabilistic nature of machine learning algorithms. Furthermore, it delves into optimisation techniques, including gradient descent and convex optimisation, empowering to optimise models and algorithms effectively.

Differences Between Artificial Intelligence, Machine Learning, and Deep Learning

Differences Between Statistical Approach, Shallow Learning, and Deep Learning

Data Types and their properties

Attribute Types

General characteristics of datasets

Data Measurement Criteria: Precision, Bias, and Accuracy

Data Pre-processing Techniques

Distance-based Dissimilarities between Datasets

Machine Learning Problems: Classification, Regression, Interpolation, andDensity Estimation

Linear Regression Model, Classification Model, and ClassificationEvaluation

Learning Algorithms: Supervised and Unsupervised

Understand and differentiate between key concepts like AI, Machine Learning, and Deep Learning. Gain a strong foundation in data properties, types, and characteristics of datasets. Furthermore, you will be able to evaluate data quality using metrics like precision, bias, and accuracy, and explore pre-processing techniques for data preparation.

Bayesian Decision Theory: Bayesian Classifier, Discriminant Functions, Minimum Error Rate Classification

Naïve Bayes Classifier

Logistic Regression Model and Parameter Estimation (Maximum-Likelihood)

Dimensionality Reduction Technique: Principal Component Analysis (PCA)

Non-parametric Techniques: K-Nearest Neighbour (KNN), Density Estimation

K-means Clustering

Decision Tree (Entropy, Gini Impurity Index)

Demonstration of All Machine Learning Algorithms

Gain proficiency in data analysis and visualisation techniques essential for extracting insights from datasets. Dive into various machine learning algorithms, including supervised, unsupervised, and reinforcement learning and tasks such as classification and regression. Understand the theoretical background of supervised methods like Linear and Logistic regression, SVM, decision trees and unsupervised methods, including clustering, KNN, and dimensionality reduction techniques (PCA).

Neurons, Perceptron Convergence Theorem, Relation Between the Perceptron and Bayes' Classifier, Batch Perceptron Algorithm, Adaptive Filtering Algorithm, Least Mean Square (LMS) Algorithm, Multilayer Perceptron, Feedforward Operation, Batch and On-line Learning, Activation Function, Backpropagation Algorithm, Rate of Learning, Stopping Criteria, XOR Problem, Loss Function, Bias and Variance, Regularization, Cross-Validation, Early-Stopping Criteria, VM, Radial Basis Function, Bagging and Boosting

Support Vector Machine (SVM)

Random Forest, Ensemble Learning, Bagging, and Boosting

Python Demo on contruction and training of neural networks for classification and regression applications

Delve into the theory and design of Artificial Neural Networks (ANNs) for classification and regression tasks, mastering essential concepts like backpropagation and stochastic gradient descent for training ANNs. Gain the necessary practical skills to implement all the algorithms using Python libraries like NumPy, pandas, scikit-learn, and Keras.

Importance of deep learning

Learning from large datasets

Types of data and architectures

End-to-end model design for feature learning and decision-making

Architecture design

Training methodology of CNN

Use cases

State-of-the-art CNN models

Python demo on object detection/image classification

Modeling of time-series data

Architecture design of RNN

Training methodology of RNN

Architectures of LSTM and advantages over RNN

Use cases

Python demo on machine translation, stock prediction

Unsupervised learning

Architecture design of AE

Convolutional AE

Training with unlabeled data

Use cases

Python demo in denoising, dimensionality reduction

Fundamentals of generative modeling

Architecture of VAE

Estimating data distribution

Training methodology of VAE

Use cases

Python demo for image generation

Generative modeling as a game-theoretic approach

Architecture design of GAN

Training methodology of GAN

Use cases Python demo on image generation, style transfer

Generative modeling through denoising

Architecture design of diffusion models

Training of diffusion models

Python demo on high-quality image generation

Attention mechanism

Advantages of Attention

Architecture design of Transformers

Training of Transformer

Python demo on language translation using Transformer

Leverage knowledge from one task to improve performance on another task

Pre-training on large datasets

Fine-tuning DL models on small dataset

Use cases

Python demo on transfer learning in computer vision

Optimisation of DL models

Transfer knowledge from a complex teacher model to a simpler student model

Training methodology for distillation

Use cases

Python demo on knowledge distillation in computer vision and natural language processing

Understanding the advantages of deep learning. Gain in-depth knowledge of deep architectures such as CNNs, RNNs, LSTMs, GRUs, Attention mechanisms, Transformers, and Autoencoders. A theoretical and practical understanding of the architectures, along with insights into design choices for better model development. Essential model training concepts like regularisation, dropout, data augmentation, batch normalisation, and hyperparameter tuning are explored for effective optimisation. Popular generative methods for AI applications such as VAEs, GANs, and Diffusion models are discussed alongside advanced topics like transfer learning, knowledge distillation, network pruning, and quantisation. Hands-on demos using TensorFlow and PyTorch on images, text, time series, language data, etc., are included for all architectures, equipping with practical skills to excel in the field of deep learning.

Industry use cases and applications of computer vision

Case studies in computer vision

Latest trends in computer vision

Latest industry use cases and applications of speech recognition

Case studies in speech recognition

Latest trends in speech recognition

VisualML Lab Pro is a no-coding machine learning laboratory platform that enables users to explore AI models through an intuitive visual environment

VisualDL Lab is a powerful no-coding deep learning laboratory platform that enables learners to explore neural networks through interactive experimentation and visual analysis

MNIST digit recognition using DNN, CNN, and SVM.

Classification of real news and fake news using decision trees.

Prediction of the iris flower species using Naive Bayes classification.

Classify photos of dogs and cats using Deep Convolutional Neural Network.

To build a movie recommender model using K-means clustering.

Identification of IoT devices using experimental radio spectrum dataset and

German Traffic Sign Recognition Benchmark: Develop a Feed Forward Neural Network and then a Convolutional Neural Network to classify between the different road signs in the dataset provided. Test it using images from the internet to validate the functioning of your model.

CIFAR-10 Object Recognition: Develop a Convolutional Neural Network to classify between the different classes of the datasets given below. Test it using images from the internet.

Sentiment analysis using Naive Bayes Classifier.

Application of deep learning in medical diagnosis using health data.

Credit card fraud detection using Random Forest Classifier.

Music recommendation system using K-NN Algorithm.

Image compression using K-means Clustering.