There will be a number of conference workshops for ICASDS 2025. For registration fee payment and other information, please consult the workshop registration information given below.

Number of seats for each workshop: 50 (on a first-come, first-served basis)

Workshop registration form:  TBA

Workshop registration fee (for a minimum of 1 and a maximum of 2):  BDT. 800 (for student); BDT. 1200 (for professional)

Workshop registration deadline: TBA

Workshop registration flyer:  TBA

Venue: ISRT lab, Institute of Statistical Research Training (ISRT), University of Dhaka, Dhaka 1000, Bangladesh

Topics and Resource Persons

Workshop-I (Session-1: 9:30 am -12:30 pm, December 27, 2025): From Business Challenge to Deployment: An Industry-Standard Framework for Applied Machine Learning Projects

Summary

Many organizations face complex operational challenges where inefficiencies and delays can result in significant financial losses. However, students graduating with strong foundational knowledge in machine learning and data science often struggle to translate these skills into actionable solutions for real-world business problems. This hands-on workshop bridges that gap by demonstrating the complete lifecycle of an industry data science project, from initial problem identification through production deployment.

Participants will learn a systematic framework for tackling operational challenges: (1) Problem Framing - translating ambiguous business concerns into well-defined, measurable objectives with clear success metrics; (2) Data Acquisition - programmatically accessing public data sources via APIs and handling real-world data quality issues including missing values, outliers, and inconsistent formats; (3) Exploratory Analysis - identifying patterns and extracting domain-specific insights that inform strategic feature engineering decisions; (4) Model Development - comparing baseline statistical approaches with advanced machine learning methods (classification and regression) while addressing practical concerns like class imbalance, temporal dependencies, and computational constraints; (5) Business Evaluation - assessing models using metrics that matter to stakeholders rather than solely academic benchmarks; and (6) Deployment Strategy - creating production-ready prediction systems with actionable recommendations and monitoring frameworks.

Through live coding demonstrations using Python and publicly available datasets, attendees will observe how organizations in developed markets leverage predictive analytics to optimize operations, reduce costs, and improve outcomes. Critically, the techniques presented are universally applicable-participants will understand how to adapt these methodologies to challenges in emerging markets including resource allocation, capacity planning, demand forecasting, and operational optimization. All code and materials will be provided for continued learning, enabling students to immediately apply these practices to problems relevant to their local contexts.

Target Audience 

Graduate students and early-career professionals with foundational knowledge of Python, pandas, and scikit-learn, Statistics, and Machine Learning concepts.

Learning Outcomes

●      Understand the gap between knowing algorithms and solving business problems

●      Master the end-to-end data science project workflow used in industry

●      Gain practical experience with API integration, feature engineering, and model deployment

●      Learn to evaluate models using business impact metrics rather than purely technical measures

●      Recognize how to adapt industry best practices to local organizational contexts

Resource Person:

Enayet Raheem, PhD

Senior Director, Advanced Analytics 

Incyte Corporation 

Wilmington, Delaware, USA

Workshop-II (Session-2: 9:30 am -12:30 pm, December 27, 2025): Hands-on Neural Network and Deep Learning 

This practical workshop offers an immersive introduction to neural networks and deep learning. Participants will explore the conceptual foundations and develop hands-on experience in designing, training, and evaluating deep learning models. The session emphasizes implementation using modern frameworks such as TensorFlow and PyTorch, with real-world examples from image and text data. 

Content 

The workshop is organized into four progressive modules, each integrating theory and coding practice. 

Part 1 – Neural Network Foundations 

Introduces the biological inspiration, mathematical foundations, and key components of neural networks. Participants will build a simple feedforward network from scratch to understand forward and backward propagation. 

Part 2 – Deep Learning Architectures 

Explores advanced structures such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs). Participants will implement both architectures using TensorFlow or PyTorch, comparing their performance on small datasets. 

Part 3 – Model Optimization and Transfer Learning 

Covers regularization (dropout, batch normalization), optimization algorithms (SGD, Adam), and transfer learning using pretrained models. Participants will fine-tune a pretrained CNN on a new dataset. 

Part 4 – End-to-End Deep Learning Project 

Participants will complete a guided project integrating all learned concepts — from data preprocessing and model design to training, evaluation, and interpretation. (45 minutes) 

Expected Outcome: Participants will gain practical experience in building and optimizing neural network and deep learning models and confidence to apply them to real-world tasks. 

Target Audience : Students and professionals with foundational knowledge of Python (or any other programming language). 

Resource Persons: 

1. Shah Mostafa Khaled, Ph.D 

Associate Professor, Institute of Information Technology, University of Dhaka 

2. Dr. Ahmedul Kabir 

Associate Professor, Institute of Information Technology, University of Dhaka 

Workshop-III (Session-3: 2:00 pm -5:00 pm, December 27, 2025): Next-Generation Statistical Computing: A Hands-On Sprint in Modern Statistical Computing 

Summary

This hands-on workshop introduces participants to modern approaches in high-performance statistical computing. The session emphasizes techniques from CPU parallelism to GPU acceleration and cloud computing. Participants will learn to develop scalable, reproducible computational strategies for statistical research and analytics.

Objectives

By the end of the workshop, participants will understand the principles of efficient computation using vectorization, parallel CPU processing, and GPU acceleration and implement them for statistical computation, extend local workflows to the Google Cloud Platform (GCP) for scalable performance.

Content

The workshop is structured into four interconnected parts designed to progressively build participants’ expertise in modern statistical computing.

·   Part 1 – Quick Foundations introduces the importance of computational efficiency in contemporary statistical analysis and demonstrates key methods, including vectorization, CPU parallelism, and GPU acceleration. Statistical techniques such as bootstrap and simulation will be used for demonstration to show improved performance and scalability. (45 minutes)

15 minutes bio break

·   In part 2 – Parallel CPU & GPU Computing. Participants take a deeper dive into parallelization, executing statistical computation for regression and cross-validation tasks in bootstrap and simulation across multi-core CPUs and GPUs, and compare performance and efficiency across environments. (30 minutes)

·    

·   Part 3 – Demonstrates the cloud deployment on GCP. This will guide participants through the setup of a GPU-enabled virtual machine on Google Cloud Platform (GCP), data management using Google Cloud Storage, and workflow execution via the RAPIDS ecosystem. (30 minutes)

15 minutes bio-break

·   Part 4 – A computation project that brings all concepts and methods together from parts 1 - 3 (45 minutes)

Expected Outcomes

Participants will leave with knowledge and skills in parallel computing methods in statistics and performing a large-scale statistical simulation.

Target Audience 

Participants should have fundamental computational skills in R and/or Python and have access to a computer and the internet during the session.

Resource Person:

Yushuf Sharker and Jaynal Abedin 

Center for Data Research and Analytics Inc., Maryland, USA 

Workshop-IV (Session-4: 2:00 pm -5:00 pm, December 27, 2025): Causal Inference for Observational Data Analysis

Summary: Causal inference focuses on identifying cause-and-effect relationships between a treatment or exposure and an outcome. Determining causal effects using randomized experiments is usually straightforward, but in observational studies, it requires additional assumptions to account for confounding and other sources of systematic bias.

In this workshop, we will introduce the fundamental concepts of causal inference grounded in the counterfactual framework. Topics will include causal Directed Acyclic Graph (DAG), the concept of causal estimands, and conditions required for their identification. Key methods such as standardization (g-formula) and propensity score-based methods (inverse probability weighting, stratification, and regression adjustment) will be briefly reviewed. Applications of these methods will be demonstrated using real-world datasets in R software.

Target Audience: Graduate students and early-career professionals with introductory knowledge of statistical modelling and basic proficiency in R programming.

Learning Outcomes:

1. Understanding the difference between causal and associational measures

2. Formulating a research question using a DAG

3. Balancing covariate distributions across treatment groups using propensity scores and inverse probability of treatment weighting

4. Estimating the causal estimand from observational data using R software

Resource Person:

Mahbub A.H.M. Latif, Tarikul Islam, and Md Muhitul Alam 

Institute of Statistical Research Training (ISRT), 

University of Dhaka, Dhaka 1000, Bangladesh