Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Introduction to Artificial Intelligence

Koundinya N V S S, Senior Data Scientist


Data Science, Artificial Intelligence and Machine Learning
 Data Analytics: systematically applying statistical or logical techniques to describe, summarize, and compare data
to derive meaningful insights and inform decision-making
 Data Science: an interdisciplinary field that uses scientific methods, processes, algorithms, and systems to
extract knowledge and insights from structured and unstructured data, applying data analysis, machine learning,
and predictive modeling to solve complex problems

Source: https://braincube.com/resource/manufacturing-ai-vs-machine-learning-vs-data-science/
How do we learn?
 Machine learning classifies as supervised, unsupervised, or reinforcement learning based on data and learning
feedback.

Source: https://www.geeksforgeeks.org/difference-between-artificial-intelligence-vs-machine-learning-vs-deep-learning/
Three classes of learning problems
 Supervised Learning: A type of machine learning where models are trained on labeled data, learning to predict
outcomes based on input-output pairs.
 Unsupervised Learning: This approach involves training models on data without labels, enabling the algorithm
to identify patterns and relationships in the data independently.
 Reinforcement Learning: A method where models learn to make decisions by performing actions in an
environment to achieve maximum reward, learning from trial and error.

Source: https://www.linkedin.com/pulse/decoding-machine-learning-supervised-unsupervised-daniel-wiczew
Machine Learning – Work Flow
 Supervised learning involves a straightforward yet powerful workflow, aimed at teaching a computer model to
make predictions or decisions, based on examples.

Input
Data

Source: https://onebelmontja.com/?e=the-workflow-of-a-machine-learning-artificial-intelligence-project-ff-kZQVzPTD
Machine Learning – Important Terms
✓ x – input, y – output (Fixed, cannot be altered)
✓ Train data and Test data
✓ Validation set, Cross Validation

✓ Activation function, Hypothesis function/Score function


✓ Loss function/Cost function
✓ Parameters Vs Hyper Parameters (weights, learning rate)
✓ Gradient Descent
✓ Forward propagation
✓ Backward propagation
Understanding the mathematics behind Machine Learning – Linear Classification
How does the learning happen? Optimisation and Parameter Update
Gradient Descent Algorithm Algorithms for parameter update
✓ SGD (Stochastic Gradient Descent)
✓ Momentum
✓ NAG (Nesterov’s Accelerated
Momentum)
✓ Adagrad
✓ Adadelta
✓ Rmsprop

Learning rate(α) hyper-parameter suggests how fast we want to go the minimum.


✓ (Batch) Gradient Descent runs through all the samples of training set before
a single parameter update
✓ Stochastic Gradient Descent, runs on only one or subset of training sample
of training set for parameter update in a particular iteration. [6]
Understanding Machine and Deep Learning

Source: https://mosaicdatascience.com/2022/02/02/how-deep-learning-facilitates-automation-innovation-and-when-to-use-it/
Neural Networks
 An Artificial Neural Network is based on a collection of connected units called artificial neurons.
 Each connection/edge (synapse) between neurons can transmit a signal (output) from one neuron to another.
 Neurons and synapses may also have a weight that varies as learning proceeds, which can increase or decrease the
strength of the signal that it sends downstream.

What happens inside a neuron?


Where comes Generative AI?
Evolution of LLMs

Source: https://medium.com/womenintechnology/ai-c3412c5aa0ac

Source: https://levelup.gitconnected.com/the-brief-history-of-large-language-
models-a-journey-from-eliza-to-gpt-4-and-google-bard-167c614af5af
How LLMs work?
 Digest vast amounts of textual data, learn from the patterns, context, and structure of language
 Predict the next word in a sequence by calculating probabilities based on the context provided by preceding
words

Source: https://www.researchgate.net/figure/Language-models-A-Language-models-are-trained-on-self-supervised-tasks-over-huge_fig1_350388949

You might also like