Deep Learning And Artificial Intelligence (AI) Training
Artificial Intelligence (AI) is the big thing in the technology field and a large number of organizations are implementing AI and the demand for professionals in AI is growing at an amazing speed. Artificial Intelligence (AI) course with ExcelR will provide a wide understanding of the concepts of Artificial Intelligence (AI) to make computer programs to solve problems and achieve goals in the world.
What Is Artificial Intelligence (AI)?
Artificial Intelligence (AI) makes computers to perform tasks such as speech recognition, decision-making and visual perception which normally requires human intelligence that aims to develop intelligent machines.
The basic grounding in the ExcelR’s practices in AI is likely to become valuable in the field of business, and profession. This course is intended to cover the concepts of Artificial Intelligence (AI) from the basics to advanced implementation.
What Are The Course Objectives?
Artificial Intelligence (AI) is becoming smarter day by day in all business functions to elevate performances. AI is used widely in gaming, media, finance, robotics, quantum science, autonomous vehicles, and medical diagnosis. AI technology is a crucial prerequisite of much of the digital transformation taking place today as organizations position themselves to capitalize on the ever-growing amount of data being generated and collected.
To build a successful career in Artificial Intelligence, this course is intended to give a complete understanding of Artificial Intelligence concepts. This course offers you get practical, hands-on experience to ensure hassle-free execution of real-life projects. This AI course leverages world-class industry expertise in making you professional data science experts.
ExcelR familiarises you with the basic terminologies, problem-solving, and learning methods of AI and also discuss the impact of AI
What Skills Will You Learn?
In this Artificial Intelligence (AI) course, you will be able to
- Understand the basics of AI and how these technologies are re-defining the AI industry
- Learn the key terminology used in AI space
- Learn major applications of AI thru use cases
Who Should Take This Course?
ExcelR’s course on Artificial Intelligence (AI) gives you the basic knowledge of Artificial Intelligence. This course doesn’t need any programming skills and best suited for:
- Well-suited for management and non-technical participants
- Students who want to learn Artificial Intelligence
Newbies who are not familiar with AI or its implications
Things You Will Learn
Description: Understand what is Artificial Intelligence and Deep Learning. Understand about the various job opportunities for AI and DL. You will learn about the most important Python libraries for building AI applications. These MUST KNOW Python libraries for Deep Learning are being upgraded on an extremely rapid pace and keeping abreast of the changes is pivotal for the success of AI experts.
- Introduction to Artificial Intelligence and Deep Learning
- Applications of AI in various industries
- Introduction to the installation of Anaconda
- Creating of Environment with stable Python version
- Introduction to TensorFlow, Keras, OpenCV, Caffe, Theano
- Installation of required libraries
Description: Understand about the various mathematical concepts which are important to learn AI implementations. These concepts will help to understand Deep Learning concepts in detail. It will also serve as a refresher for learning various Neural Network algorithms, which are synonymous to Deep Learning.
- Introduction to Data Optimization
- Calculus and Derivatives Primer
- Finding Maxima and Minima using Derivatives in Data Optimization
- Data Optimization in Minimizing errors in Linear Regression
- Gradient Descent Optimization
- Linear Algebra Primer
- Probability Primer
Description: Understand about the basics of the first algorithm – Perceptron Algorithm, its drawbacks and how we can overcome those challenges using Artificial Neural Network or Multilayer Perceptron Algorithm. The various activation functions will be understood in detail and practical exposure to R programming and Python programming is the highlight of this module.
- Understand the history of Neural Networks
- Learn about Perceptron algorithm
- Understand about Backpropagation Algorithm to update weight
- Drawbacks of Perceptron Algorithm
- Introduction to Artificial Neural Networks or Multilayer Perceptron
- Manual calculation of updating weights of final layer and hidden layers in MLP
- Understanding of various Activation Functions
- R code and Python code to understand about practical model building using MNIST dataset
Description: Learn about the various Error functions, which are also called Cost functions or Loss functions. Also, understand about the entropy and its use in measuring error. Understand the various optimization techniques, drawbacks and ways to overcome the same. This you will learn alongside various terms in implementing neural networks.
- Understand about challenges in Gradient
- Introduction to various Error, Cost, Loss functions
- ME, MAD, MSE, RMSE, MPE, MAPE, Entropy, Cross Entropy
- Vanishing / Exploding Gradient
- Learning Rate (Eta), Decay Parameter, Iteration, Epoch
- Variants of Gradient Descent
- Batch Gradient Descent (BGD)
- Stochastic Gradient Descent (SGD)
- Mini-batch Stochastic Gradient Descent (Mini-batch SGD)
- Techniques to overcome challenges of Mini-batch SGD
- Nesterov Momentum
- Adagrad (Adaptive Gradient Learning)
- Adadelta (Adaptive Delta)
- RMSProp (Root Mean Squared Propagated)
- Adam (Adaptive Moment Estimation)
Description: Learn about practical applications of MLP when output variable is continuous and discrete in two categories and multi-category. Understand also about handling balanced vs imbalanced datasets. Learn about techniques to avoid overfitting and various weight initialization techniques.
- Binary classification problem using MLP on IMDB dataset
- Multi-class classification problem using MLP on Reuters dataset
- Regression problem using MLP on Boston Housing dataset
- Types of Machine Learning outcomes – Self-supervised, Reinforcement Learning, etc.
- Handling imbalanced datasets and avoiding overfitting and underfitting
- Simple hold-out validation
- K-Fold validation
- Iterated K-fold validation with shuffling
- Adding weight regularization
- L1 regularization
- L2 regularization
- Drop Out and Drop Connect
- Early Stopping
- Adding Noise – Data Noise, Label Noise, Gradient Noise
- Batch Normalization
- Data Augmentation
- Weight initialization techniques
- Xavier, Glorot, Caffe, He
Description: Though CNN has replaced most of the computer vision and image processing concepts, a few application require the knowledge of Computer vision. We will learn about the application using the defacto library OpenCV for image processing. How to build machine learning models when we have limited data is explained as part of this module.
- Understanding about Computer Vision related applications
- Various challenges in handling Images and Videos
- Images to Pixel using Gray Scale and Color images
- Color Spaces – RGB, YUV, HSV
- Image Transformations – Affine, Projective, Image Warping
- Image Operations – Point, Local, Global
- Image Translation, Rotation, Scaling
- Image Filtering – Linear Filtering, Non-Linear Filtering, Sharpening Filters
- Smoothing / Blurring Filters – Mean / Average Filters, Gaussian Filters
- Embossing, Erosion, Dilation
- Convolution vs Cross-correlation
- Boundary Effects, Padding – Zero, Wrap, Clamp, Mirror
- Template Matching and Orientation of image
- Edge Detection Filters – Sobel, Laplacian, LoG (Laplacian of Gaussian)
- Bilateral Filters
- Canny Edge Detector, Non-maximum Suppression, Hysteresis Thresholding
- Image Sampling – Sub-sampling, Down-sampling
- Aliasing, Nyquist rate, Image pyramid
- Image Up-sampling, Interpolation – Linear, Bilinear, Cubic
- Detecting Face and eyes in the Video
- Identifying the interest points, key points
- Identifying corner points using Harris and Shi-Tomasi Corner Detector
- Interest point detector algorithms
- Scale-invariant feature transform (SIFT)
- Speeded-up robust features (SURF)
- Features from accelerated segment test (FAST)
- Binary robust independent elementary features (BRIEF)
- Oriented FAST and Rotated Brief (ORB)
- Reducing the size of images using Seam Carving
- Contour Analysis, Shape Matching and Image segmentation
- Object Tracking, Object Recognition
Description: Understand about the various layers of CNN and understand how to build the CNN model from scratch as well as how to leverage upon the CNN model which is pre-trained. Understand about the best practices in building CNN algorithm and variants of convolution neural network.
- Understand about various Image related applications
- Understanding about Convolution Layer and Max-Pooling
- Practical application when we have small data
- Building the Convolution Network
- Pre-processing the data and Performing Data Augmentation
- Using pre-trained ConvNet models rather than building from scratch
- Feature Extraction with and without Data Augmentation
- How to Visualize the outputs of the various Hidden Layers
- How to Visualize the activation layer outputs and heatmaps
Description: Understand about how to deal with sequence data including textual data as well as time series data and audio processing. Understand about advanced RNN variant models including LSTM algorithm and GRU algorithm. Also learn about bidirectional RNN, LSTM and deep bidirectional RNN and LSTM. Learn about various unstructured textual pre-processing techniques.
- Understand about textual data
- Pre-processing data using words and characters
- Perform word embeddings by incorporating the embedding layer
- How to use pre-trained word embeddings
- Introduction to RNNs – Recurrent layers
- Understanding LSTM and GRU networks and associated layers
- Hands-on use case using RNN, LSTM, and GRU
- Recurrent dropout, Stacking recurrent layers, Bidirectional recurrent layers
- Solving forecasting problem using RNN
- Processing sequential data using ConvNets rather than RNN (1D CNN)
- Building models by combining CNN and RNN
Description: Understand about unsupervised learning algorithm such as GAN as well as Autoencoders. GANs are used extensively in artificially generating speech, images which can be used in computer games. Deep Dream is such an algorithm which using GAN to generate images. Autoencoders will take input as an image and traverse through the network and then regenerates the same image. Learn about how these intermediate layer representations can be used in other neural network deep learning models.
- Text generation using LSTM and generative recurrent networks
- Understanding about DeepDream algorithm
- Image generation using variational autoencoders
- GANs theory and practical models
- The Generator, the Discriminator, the Adversarial network
- Deep Convolution Generative Adversarial networks
- Producing audio using GAN
- Unsupervised learning using Autoencoders
Description: Reinforcement learning is majorly used in AI-based games. Q-learning is one such Reinforcement machine learning algorithm which is using in game theory. Finally, any of the ongoing Kaggle competition will be the prime focus and to be in the top 100 will be of prime importance. This will bring optimal visibility of the profiles to the employers and participants can be directly hired.
- Exploration and Exploitation
- Experience Replay
- Model Ensembling
- Final project using a live Kaggle competition