ArticlesVideosEventsResearch PapersAI WikiPodcastsPodcastsWhite papersTutorials

AI Wiki

Info-nuggets to help understand Machine Learning and AI concepts. Explore MLOps, Synthetic data, and their significance throughout the ML lifecycle.

Thank you! Latest edition is in your way :)
Oops! Something went wrong while submitting the form. Please try again.

AI Regulations in China

AI Regulations in the European Union (EU)

AI Regulations in the US

AI Regulations in India

Model safety

Synthetic & Generative AI

MLOps

Model Performance

ML Monitoring

Explainable AI

Is Explainability critical for your AI solutions?

Schedule a demo with our team to understand how AryaXAI can make your mission-critical 'AI' acceptable and aligned with all your stakeholders.

Book a Demo

Privacy Evaluation

Assessing how well a model or dataset protects sensitive information

F-Score (F1-Score)

A measure used to evaluate the performance of a classification model

Constant Features

Features in a dataset that have the same value for all observations

High Feature Correlation

A situation where two or more input features in a dataset exhibit a strong linear relationship

Target Drift

Changes in the distribution of the target variable (output) over time, which can affect the performance of machine learning models

Stochastic Gradient Descent (SGD)

Optimization algorithm used primarily for training machine learning models

RandomForest

An ensemble learning algorithm used primarily for classification and regression tasks

CatBoost (Categorical Boosting)

A powerful gradient boosting algorithm specifically designed to handle categorical data more effectively

LightGBM (Light Gradient Boosting Machine)

A gradient boosting framework that uses decision trees as its base learners

XGBoost (eXtreme Gradient Boosting)

Advanced machine learning algorithm based on the gradient boosting framework

CTGAN (Conditional Tabular Generative Adversarial Network)

Deep learning-based model designed for generating synthetic tabular data

GPT-2 (Generative Pre-trained Transformer 2)

Powerful and widely used language model

Internet Information Service Algorithm Recommendation Management Regulations

Govern the use of recommendation algorithms across various sectors

Generative AI Measures in China

Apply to providers of generative AI services and users of these services

Provisions on the Administration of Deep Synthesis of Internet-based Information Services

Regulations regarding handling deep synthesis data and technologie

Artificial Intelligence and Algorithmic Fairness Initiative

Initiated by the the U.S. Equal Employment Opportunity Commission (EEOC)

The EU AI Act

The world's first comprehensive law on artificial intelligence (AI)

Artificial Intelligence Risk Management Framework (AI RMF 1.0)

Managing risks and promoting trustworthy AI systems in organizations

Federal Trade Commission (FTC)

Tackling deceptive or unfair advertising claims related to AI

President Biden's Executive Order on AI

Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence

Principles for Responsible AI

First part of an approach paper by NITI Aayog

Digital India Act

Establishing a comprehensive legal framework for India's digital economy

Draft National Data Governance Framework Policy

To modernize and improve the government's data collection and management procedures

National Strategy for Artificial Intelligence #AIFORALL: NITI Aayog

India's first national strategy for AI

National Cybersecurity Reference Framework

Cybersecurity framework to guide critical sectors

Global Partnership on Artificial Intelligence (GPAI)

India became a member of the Global Partnership on Artificial Intelligence (GPAI)

Top-k

Constrains the model's predictions during the generation process to the top-k most probable tokens at each step

Temperature

Parameter often associated with models that generate probability distributions

Low-Rank Adaptation (LoRA)

Approach for fine-tuning large-scale, pre-trained models

Quantization

Process of converting continuous infinite input values from a large set to discrete finite output values in a smaller set

Hallucination

Model output generated that is not grounded in the input data but instead is generated imaginatively or erroneously

Multi-modal models

Capable of understanding and generating content across multiple data types or ‘modalities’

Mixture of experts (MoEs)

Learning technique used in generative AI to enhance the capabilities of models

Mamba

Innovative state-space model (SSM) designed to handle complex, data-intensive sequences

Opensource vs. Closed Source Models

The choice between open-source and closed-source models has emerged as a key debate within generative AI

Tokens

Smallest unit or chunks of text that a model processes

Diffusion Models

Models designed to generate realistic, high-resolution images of varying quality.

Transformers Models

Foundational blocks for various natural language processing (NLP) and generative AI tasks

Model safety

Measures taken to ensure that AI is deployed in ways that do not harm humanity

Vector database

Category of databases which store data as numerical representations of various complex forms of data

Large Language Models (LLMs)

Machine learning models that utilize deep learning algorithms for understanding and analyzing natural language

Foundation models

Serve as a basis for a wide range of natural language processing (NLP), computer vision tasks, etc.

Synthetic AI

Generating synthetic data that imitate real-world data

Generative AI

Generate new data samples that resemble the patterns or distribution of the input data

Z-Test

Z-test is a statistically significant test used to determine whether two population means are different from each other.

Chi-square test

The chi-square test is a statistical hypothesis test used to determine if there is a significant discrepancy between the expected and observed frequencies in one or more categories. 

Kolmogorov–Smirnov test (K–S test or KS test)

The Kolmogorov-Smirnov statistic determines the probability of two samples coming from the same distribution. It does not make any assumptions about the underlying distribution of the data, making it a nonparametric test. 

Wasserstein distance

The Wasserstein distance is a distance function for measuring the distance between two probability distributions.

Jensen-Shannon(JS) Divergence

The Jensen-Shannon distance measures the similarity between two probability distributions.

Population Stability Index (PSI)

PSI is a measure of the degree to which the distribution of a population has shifted over time or between two different samples of a population in a single number.

Kullback-Leibler (KL) divergence

KL divergence, also referred to as relative entropy, calculates the difference between two probability distributions.

ML Audit

ML audit ensures identifying associated risks and developing safeguard controls to avoid risks.

Bias Monitoring

ML bias is a phenomenon where some aspects of datasets with equal significance are given more weight or representation than others, leading to skewed outcomes.

Data Monitoring

A practice of proactively examining and assessing crucial business data against quality control guidelines to maintain a high, consistent standard of data quality proactively

ML Observability

ML observability entails tracking the performance of ML system across its lifecycle, right from when it is being built, to pre and post-production.

MLOps

MLOps is the discipline of deploying and maintaining machine learning models in production efficiently and reliably.

Regression model

A statistical technique for analyzing the correlation between independent features or variables and a dependent feature, result, or variable.

Classification model

Classification algorithm is a type of supervised learning that uses machine learning algorithms to learn how to categorize instances based on training data.

Model confidence score

Model confidence score expresses how confident the machine learning model or NLP service is that the relevant intent has been accurately assigned

Model threshold

Thresholding is a technique for getting the most value out of a machine learning classifier.

Surrogate models

An interpretable model that is trained to approximate the predictions of a black box model.

What-if

Enables teams to run hypothetical scenarios to understand what features need to be tweaked to obtain desired predictions

Similar Cases

Similar cases help machine learning models discover groups of similar examples within the data.

Feature Importance Store

Feature importance relates to methods for scoring each input feature for a particular model; the scores indicate the importance of each feature.

LOCO

The LOCO method seeks to understand the significance of a particular feature for a model's prediction performance.

SHAP

SHAP (SHapley Additive exPlanations) is a game theoretic approach for explaining the output of any machine learning model.

LIME

LIME approach attempts to comprehend the model by varying the input of data samples and observing how the predictions change.

Explainability

AI explainability (XAI) refers to the process of explaining to an individual the decision-making process of an AI model

Fairness/ Bias Monitoring

Bias is the situation where the model consistently predicts distorted results because of incorrect assumptions.

ML Monitoring

ML monitoring encompasses establishing alerts on key model performance metrics such as accuracy and drift.

Accuracy

Machine learning model accuracy is the measurement used to discover which model is best at recognising correlations and patterns between variables in a dataset.

Data Processing

Data processing involves transforming raw data into something meaningful that can be used by a machine learning model.

Precision

Precision is defined as the percentage of positive or relevant outcomes.

Recall/ Sensitivity or True Positive Rate

The percentage of actual positives that were accurately identified is known as Recall.

Specificity/ True Negative Rate:

The percentage of true negatives the model correctly identifies is known as specificity.

Precision-recall curve

The precision-recall curve depicts the tradeoff between precision and recalls for various thresholds.

Confusion Matrix

The performance of the classification models for a given set of test data is evaluated using a confusion matrix.

F score

The harmonic mean of precision and recall is used to calculate the F-measure, giving each the same weight.

ROC Curves and ROC AUC

ROC curve is a plot summarizing how well a binary classification model performs for the positive class.

Data Drift

Data drift is the change in model input data that causes a decline in model performance.

Model Drift

Model drift transpires when the ML models predictive ability "drifts" from the performance over the training period.

AryaXAI provides the most accurate explainability and alignment stack to deliver accurate, true-to-model explainability, monitoring, risk management, and alignment techniques essential for highly mission-critical or regulated AI solutions.

Company
Resources
Products

Follow Us

Follow Us

hello@aryaxai.com

Subscribe

Stay up to date with all updates

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.

© Copyright 2024, Lithasa Technologies Pvt. Ltd,
Address: CoWrks, 3rd Floor, Prudential Building,
Powai, Mumbai- 400076

Terms and ConditionsPrivacy PolicyPayments and Refunds PolicyContent Removal