Scribd
Upload
37 views29 pages

Class 12 Model Lifecycle AI 843

Uploaded by

admin
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
37 views29 pages

Class 12 Model Lifecycle AI 843

Uploaded by

admin
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 29

AI MODEL LIFECYCLE

Stage I – Project Scoping

Stage II – Design & develop

Stage III – Test & Deploy


BUILDING AN AI SYSTEM

• Creating an Artificial Intelligence (AI) system involves recreating human


skills in a machine in a way that its computing power outperforms
humans.
BUILDING AN AI SYSTEM

• Creating an Artificial Intelligence (AI) system involves recreating human


skills in a machine in a way that its computing power outperforms
humans.

• Without a robust system, a business could potentially fail, as errors could


tarnish its reputation.
AI MODEL LIFECYCLE
Helps in creating a robust system with high efficiency

• The concept of 'AI Model Lifecycle Management' helps manage this


complex workflow, ensuring the desired results

• The AI model lifecycle is a repeated sequence of steps


AI MODEL LIFECYCLE
It is a cyclical process
STAGE I – PROJECT PLANNING AND DATA COLLECTION
Explore the reasons for implementing artificial intelligence (AI) in operations.

• Estimate the amount of data needed for the project.

• Assign responsibility for data collection.

• Develop a strategy to successfully complete the project.


STAGE I – PROJECT PLANNING AND DATA COLLECTION
Explore the reasons for implementing artificial intelligence (AI) in operations.

• Consideration of whether a simpler solution could address the


problem or if AI is truly needed.

• Defining success metrics, and anticipating key resources and


steps are key parts of this phase.
STAGE II – DESIGN AND TRAINING OF ML MODEL
It is important that the data for Al solution be accurate and relevant

• Using poor quality data can extend project timelines and lead to
inaccurate results.

• Data can be obtained from primary sources like company's own


CRM or IoT devices, or secondary sources such as third-party
data providers or government publications.
STAGE II – DESIGN AND TRAINING OF ML MODEL
A. Data Labelling

• Data labeling involves detecting and tagging data samples.

• This helps Machine Learning (ML) models understand and interpret


digital data.

• Data labeling can be done manually or with special software.


STAGE II – DESIGN AND TRAINING OF ML MODEL
B. AI Modelling
STAGE II – DESIGN AND TRAINING OF ML MODEL
B. AI Modelling

1. Binary classification model: As the name implies, ML models for


binary classification problems predict a binary outcome.

For example:

• Is this e-mail spam or not (yes or no)?


• Is this app review written by a real person or a bot?
STAGE II – DESIGN AND TRAINING OF ML MODEL
B. AI Modelling

2. Multi-class classification model: In this case, models predict an


outcome into one of the three or more classes.

For example:

• Is a child holding a toy, a book, or a pen?


• Is this genre of music rock, jazz or hip hop?
STAGE II – DESIGN AND TRAINING OF ML MODEL
B. AI Modelling

3. Regression model: ML models for regression problems predict the


relationship between a single dependent variable and one or more
independent variables.

For example:

• How much energy will be utilized in Delhi tomorrow?


• How many chairs will be sold this month?
STAGE II – DESIGN AND TRAINING OF ML MODEL
C. Training the Model
STAGE II – DESIGN AND TRAINING OF ML MODEL
C. Training the Model

• We input data into our ML (Machine Learning) algorithms.

• Initially, the model's accuracy might be low, but this improves with more training.

• After training, evaluating, and testing the model, we move to the final stage.
STAGE II – DESIGN AND TRAINING OF ML MODEL
C. Training the Model
• Languages: Python is most popular, R, Scala

• Open Frameworks: Scikit-learn, XGBoost, TensorFlow, etc.

• Approaches and Techniques: Classic ML techniques from regression


all the way to state-of-the-art GANs (General Adaptive neural
network) and RL (Reinforcement learning).

• Productivity-enhancing Capabilities: Visual modelling, Auto Al to help


with feature engineering, algorithms Selection and hyper parameter
optimization.

• Development Tools: Data Robot, H;O, Watson Studio, Azure ML


Studio, Sagemaker, Anaconda, etc.
STAGE II – DESIGN AND TRAINING OF ML MODEL
C. Training the Model

• Productivity-enhancing Capabilities: Visual modelling, Auto Al to help


with feature engineering, algorithms Selection and hyper parameter
optimization.

• Development Tools: Data Robot, H;O, Watson Studio, Azure ML


Studio, Sagemaker, Anaconda, etc.
STAGE III – DEPLOYMENT AND MAINTENANCE
Evaluation and Testing

• AI models need to be tested to ensure they work with new data.

• Continuous monitoring and adjustments are necessary after the AI is live.

• Specific test cases should be created to validate machine learning models.


STAGE III – DEPLOYMENT AND MAINTENANCE
Deployment

• The AI model is set up at the client's place.

• where to store data, who will handle problems if the system


breaks down, who will be in charge of keeping the system
running and making it better.
STAGE III – DEPLOYMENT AND MAINTENANCE
AI model Tuning

• New data is constantly given to the AI model to continuously improve its


function.

• Check that the Al system is still meeting the established business goals
from time to time.
EXPECTED QUESTIONS
What is the need of Al model lifecycle management?

• In enterprise, the critical role of Al requires a well-defined and strong


methodology and platform. In the absence of strong methodology and
platform, a business might fail.

• For example, trying to detect a wrong transaction among the various


other bank transaction, model might make mistakes and it would stain
the name of the business.

• In the long pipeline for Al, response time, quality and other elements
should be dealt with utmost caution.

• Therefore, what we call ’Al Model Lifecycle Management manages the


complicated Al pipeline and helps ensure the necessary results in
enterprise.
EXPECTED QUESTIONS
How many stages are there in Al lifecycle?

• The Al lifecycle involves various stages, from data collection, data


analysis, feature engineering and algorithm selection to model
building, tuning, testing, deployment, management, monitoring and
feedback loops for continuous improvement.

• Generally, every Al project lifecycle encompasses three main stages:

▪ Project scoping
▪ Design or build phase
▪ Deployment in production.
EXPECTED QUESTIONS
What main points are covered in design phase ?

• The Design phase is essentially an iterative process comprising all the


steps relevant to building the Al or machine learning model: data
acquisition, exploration, preparation, cleaning, feature engineering,
testing and running a set of models

• Aim is to predict behaviours or discover insights in the data.

• From self-driving cars to facial emotion recognition, data is at the


core of all Al projects. Artificial Intelligence systems are capable of
recognizing patterns and making decisions based on the statistical
models.
EXPECTED QUESTIONS
What are the two sources of data in data collection phase?

1. Primary Sources: Primary sources are sources that provide data that
originates from your own company. You can acquire such data from
tools like CRM or IoT devices.

2. Secondary Sources: Secondary sources refer to external sources that


have relevant data of interest to you. Those can be Third-Party data
providers or government publications.
EXPECTED QUESTIONS
Name the popular Al Algorithms.

The popular Al algorithms are:

• Neural networks

• Decision trees

• Random forest

• Logistic regression
What Next?
• Storytelling Through data

• Capstone Project

You might also like