eChapters

The rapid growth of Large Language Models has produced remarkable advances in artificial intelligence but has also increased computational requirements, energy consumption and environmental impact. This chapter examines the relationship between large-scale AI development, energy efficiency and sustainable computing. Modern AI models may contain billions or even trillions of parameters and require extensive datasets, powerful accelerators and energy-intensive data-centre infrastructure for training and inference. The chapter investigates the environmental consequences associated with model development, including energy consumption, carbon emissions, cooling requirements and resource utilization. It examines model-level optimization techniques such as pruning, quantization and knowledge distillation as methods for reducing computational demands. System-level approaches, including specialized hardware accelerators, energy-efficient data centres, improved cooling systems, renewable energy and intelligent workload scheduling, are also considered. The distinction between energy efficiency and broader sustainability is emphasized, with attention to metrics such as energy per inference, total energy consumption, carbon emissions and hardware lifecycle impacts. The chapter further highlights the importance of transparent sustainability reporting, green computing and governance policies. Sustainable AI development requires balancing model performance and innovation with responsible resource use and long-term environmental objectives

The integration of Large Language Models into software development has created powerful capabilities for automated code generation, debugging, testing and application development. However, AI-generated code can introduce security vulnerabilities when models produce incorrect, unsafe or poorly validated programming solutions. This chapter examines the need for security-aware frameworks that combine LLM-based code generation with verification, rule-based evaluation and risk management. It considers the limitations of conventional rule-based systems, including rule explosion, inflexibility, maintenance overhead and limited personalization, and explores how AI-driven approaches can improve adaptability while maintaining safeguards. Particular attention is given to secure integration in sensitive domains such as healthcare, where generated code and automated analytical systems may influence clinical information and decision-making. The chapter considers data-driven evaluation, machine learning, genomic information and risk detection as components of secure AI systems. It emphasizes the importance of validating generated code rather than treating model output as automatically reliable. Secure deployment requires testing, human oversight, access controls, verification mechanisms and continuous monitoring. The chapter therefore presents safe LLM integration as a multidisciplinary challenge combining AI capability, software engineering, cybersecurity and responsible governance.

AI-enabled livestock technologies can improve productivity, health monitoring, animal welfare and environmental performance, but they also raise significant ethical and governance concerns. This chapter examines data ownership, privacy, algorithmic bias, transparency, accountability, cybersecurity, socioeconomic inequality and animal well-being. Precision livestock systems continuously monitor animals through sensors, computer vision, machine learning and IoT networks, enabling preventive decisions and automated interventions. However, excessive dependence on algorithms may reduce human oversight and treat animals merely as data-producing units. The chapter advocates an interdisciplinary framework for responsible AI that aligns innovation with animal-welfare standards, farmer interests, social values, equitable access, regulatory requirements and long-term environmental sustainability.

Agriculture has progressed from labour-intensive traditional practices through mechanization and the Green Revolution to precision agriculture and Agriculture 5.0. This chapter traces that historical development and identifies limitations associated with conventional farming, including inefficient resource use, environmental degradation and weak adaptability. It explains how GPS-guided machinery, sensors and precision technologies prepared the foundation for intelligent systems. Artificial neural networks, deep learning and fuzzy logic now process data from IoT devices, UAVs and satellites to support automated irrigation, crop monitoring and soil analysis. The chapter emphasizes that integrating human knowledge with computational power is increasingly essential for global food security, sustainability and climate resilience.

Modern farm planning involves nonlinear, discontinuous and multi-objective problems that conventional optimization methods may not solve effectively. This chapter examines Genetic Algorithms, Particle Swarm Optimization and Differential Evolution for improving agricultural resource allocation. These nature-inspired techniques can simultaneously address crop yield, water use, fertilizer application, environmental effects and financial viability. Their integration with deep-learning forecasts and crop-simulation systems enables more realistic optimization. Multi-objective evolutionary algorithms identify Pareto-optimal alternatives, allowing decision-makers to balance productivity and ecological sustainability. The chapter also emphasizes explainable artificial intelligence, economical implementation and farmer usability as necessary conditions for transferring evolutionary optimization methods from experimental research to practical farm management.

Advanced artificial-intelligence models can achieve high predictive performance, but their black-box nature limits clinical trust and accountability. This chapter examines interpretability and explainability as essential requirements for trustworthy healthcare analytics. It reviews intrinsically interpretable models, post-hoc explanation methods, example-based approaches and network-based explanations. A framework combining data preparation, model selection, ensemble learning, calibration, uncertainty estimation and explanation validation is presented. An explainable sepsis-risk prediction case study demonstrates logistic modelling, ensemble methods, Bayesian uncertainty, SHAP attribution and network analysis. The chapter also addresses causal reasoning, adversarial robustness, temporal and multimodal data, human-centred explainability, ethical responsibilities and regulatory requirements for dependable clinical AI.

Factorial experiments provide an efficient method for studying two or more factors simultaneously at different levels. This chapter introduces the concept and notation of factorial arrangements and explains treatment combinations in experiments such as 2 × 2 and higher-order factorials. It distinguishes among simple effects, main effects, and interaction effects, emphasizing the importance of interactions in interpreting biological responses. Factorial experiments allow researchers to examine the effects of individual factors as well as their combined influence. The chapter also explains that factorial treatments may be arranged using different basic experimental designs, such as Completely Randomized, Randomized Block, or Latin Square designs, according to experimental objectives and conditions.

Abstract

FASTA, which stands for “Fast All”, is a sequence alignment tool developed by William R. Pearson and David J. Lipman in 1985. It is commonly used to identify similarities between DNA and protein sequences. FASTA employs a “hashing” strategy to locate matches for short stretches of identical residues with a length of k, referred to as k-tuples or k-tups. This method enables FASTA to quickly search large databases for sequence similarities, making it a popular tool for identifying homologous sequences, predicting gene functions, and analyzing evolutionary relationships. This chapter elaborates some basics of FASTA.

Healthcare organizations require collaborative learning methods that protect sensitive patient information. This chapter introduces federated learning as a framework for training models across decentralized institutions without transferring raw clinical data. Horizontal, vertical and hybrid federated architectures are examined alongside communication protocols, threat models and security requirements. Privacy-enhancing methods—including differential privacy, secure aggregation, cryptographic techniques and trusted execution environments—are discussed. Applications in medical imaging, electronic health-record analysis, wearable devices and remote patient monitoring demonstrate the framework’s practical value. The chapter also evaluates accuracy, privacy, communication cost, scalability and robustness while addressing data heterogeneity, fairness, explainability, legal compliance and ethical challenges in collaborative healthcare analytics.

Finger millet (Ragi/ Mandua) is the principal small millet grown in South Asia. It is believed to be native of India (De condole, 1886) but according to Vavilov (1926) it originated in Abyssinia.

12.1 Meaning and Definition

A flow chart is a graphical or symbolic representation of a process. Each step in the process is represented by a different symbol and contains a short description of the process step. The flow chart symbols are linked together with arrows showing the process flow direction. This diagrammatic representation can give a step-by-step solution to a given problem. Flowcharts are used in analyzing, designing, documenting or managing a process or program in various fields.

Introduction
 
Arid western part of Rajasthan which contains the Indian segment of the Thar Desert, accounts for ~62% of the country’s hot arid area and is a region of low rainfall. Under such physico-climatic situation, fluvial processes function much less frequently and more episodically than the aeolian processes, but their impacts on landforms and society are large, as noticed on several occasions in arid regions. There is a distinct east-west gradient of rainfall and temperature in the region. Along the eastern margin bordering Aravallies, the mean annual rainfall is 500 mm while in the westernmost part at Jaisalmer, it is less than 150 mm. The rainfall is largely monsoon driven which brings 80 - 90 per cent of total rainfall received during the year. There is large inter-annual rainfall variability of rainfall which leads to climatic extremes of drought and floods of different magnitudes. For example on an average, the Thar Desert in India experiences flood-causing rainfall in 2-3 years in a decade, but a maximum of 5 flood causing rainfall years in a decade was recorded during the 1950s, 1970s and the 1990s (Rao et al. 2006).

Cowpea is a quick growing leguminous forage crop. It is usually grown mixed with cereal fodders and grasses to improve the nutritive value of the herbage. It contains 20 – 24 % crude protein, 43 - 49 % neutral detergent fibre, 34 – 37 % acid detergent fibre, 23– 25 % cellulose and 5 – 6 % hemicelluloses on dry matter basis. The digestibility of cowpea fodder is above 70%. Cowpea can be grown under partial shaded conditions. It is an excellent cover crop, which suppresses weeds and enriches the soil. Cowpea requires warm climate with good atmospheric humidity. It can be grown in kharif as well as in zaid season.

Forages: The term forages denote plants that are used for feeding the animals that are cultivated in the farm, or it may be a wild used for stock feeding grown for the vegetative purpose. The vegetative phase is encouraged and the reproductive phase is suppressed.

Introduction

As India is the leading global economy with livestock contributing 4.66% in total GVA to Indian Economy (PIB, 2024) and ranks 1st in Milk Production, 2nd in Egg production, 5th in meat production (BAHS, 2024). Being the largest exporter of milk and meat in the world , this has led to the intensification of animal production due to the increasing demand of major livestock products (MLPs), which has led to an increasing overall use of antimicrobials. Antimicrobials are invariably used to preserve animal and public health, but also as growth promoters at a sub-therapeutic level. The overuse and misuse of antibiotics can contribute to antimicrobial resistance. The growing resistance to antimicrobial drugs could reverse the therapeutic benefits; animals’ resistance to antimicrobial drugs, reduces productivity and leads to economic losses. Unwanted antimicrobials residues may be present in products of animal origins, in animal waste contaminating soil and water and the environment. Seventy-five to ninety percent of antimicrobials used in livestock are excreted, mostly unmetabolized. AMR poses a serious threat to the safety and quality of animal origin food, food security and livelihoods. Unhealthy and unproductive animals are no longer able to generate food products of acceptable safety and quality for human consumption and can no longer contribute to income generation. Insufficient standards for residues in animal food products, reduces the livestock sectors potential of access to trade and increases public health risks (https://www.fao.org/antimicrobial-resistance/key-sectors/animal-production/ en/).

Foundation architectures provide the technological basis for modern artificial intelligence systems and enable models to generalize across diverse tasks and modalities. This chapter examines the evolution of neural architectures that ultimately led to the dominance of the Transformer in contemporary language and multimodal AI. It begins with early neural networks, convolutional neural networks and recurrent neural networks before introducing attention mechanisms and the Transformer architecture. Particular attention is given to the major components of Transformers, including self-attention, multi-head attention, positional encoding and feedforward layers. Encoder, decoder and encoder–decoder configurations are examined through influential models such as BERT and GPT. The chapter further considers Vision Transformers and CLIP, demonstrating the extension of Transformer principles beyond natural language processing. At the same time, limitations such as computational complexity, memory requirements and long-context processing are discussed. Alternative approaches, including state-space models such as Mamba, efficient attention mechanisms, hierarchical architectures and Mixture-of-Experts models, are also explored as promising directions for scalable, efficient and adaptive AI.

Labelling is a critical component of food regulation, consumer communication, traceability, and brand credibility. This chapter examines the regulatory requirements governing ice cream and frozen dessert labelling in India, with emphasis on FSSAI standards and declarations. It explains the distinction between product standards and label declarations and introduces systems for managing recipe masters, nutrition information, label copy, artwork, and packaging approvals. Mandatory declarations, ingredient information, allergen declarations, nutritional information, claims, and common labelling errors are considered. The chapter also identifies departmental responsibilities and best practices for label compliance. Proper regulatory control ensures accurate consumer information, reduces legal risk, and supports consistent packaging and commercial operations.

4.1 Introduction to Function in Python

Python Functions is a block of related statements designed to perform a computational, logical, or evaluative task. The idea is to put some commonly or repeatedly done tasks together and make a function so that instead of writing the same code again and again for different inputs, we can do the function calls to reuse code contained in it over and over again. Functions can be both built-in or user-defined.

Function means sub-program

1. Function definition

2. Function call

Example:

def message: #function definition

print(?hello python”)

message() #function call

Artificial intelligence and machine learning provide the technological foundation for intelligent agricultural applications. This chapter introduces supervised, unsupervised and reinforcement learning, model-development pipelines, feature engineering and evaluation metrics. It explains how algorithms can be selected and implemented for agricultural datasets that may be noisy, incomplete, variable or weakly labelled. Advanced topics include deep-learning architectures, transfer learning and model optimization. Practical considerations involving preprocessing, interpretability and deployment constraints are also discussed. Agricultural case studies illustrate AI applications in crop-yield prediction, plant-disease detection and environmental modelling. The chapter provides the technical foundation needed to design, evaluate and responsibly implement AI-driven agricultural solutions.

Artificial intelligence offers practical solutions to agricultural problems associated with population growth, food demand, climate change and limited natural resources. This chapter introduces machine learning, deep learning, computer vision and natural language processing, together with enabling technologies such as IoT, remote sensing and cloud computing. It explains their applications in crop monitoring, disease recognition, yield forecasting, irrigation, soil assessment and livestock management. AI can improve decision-making, resource efficiency, automation and precision farming while reducing reliance on manual labour. The chapter also examines adoption barriers, including high costs, limited digital literacy, data-security concerns and weak rural infrastructure, and considers future developments in autonomous and climate-resilient farming.