ChatGPT, LLMs and Deep Learning with Python

Subject Category: ICT - Information & Communication Technologies

Start date: 01/02/2025

Program Information

PROGRAM CODE: 1701
REGISTRATION PERIOD: 17/12/2024 - 14/02/2025
COURSE PERIOD: 01/02/2025 - 28/02/2025
DURATION: 1 month
TRAINING HOURS: 100.00
CREDITS ECTS: 4.00

Scientific Responsibility: SYMEONIDIS PANAGIOTIS

Academic Responsibility: SYMEONIDIS PANAGIOTIS

INITIAL COST: 120,00 € (Program Discounts)

DESCRIPTION

OUTLINE

CONTACT

DISCOUNTS - TERMS OF USE

Program Description

Subject & purpose of the training program:

This educational program presents in detail the book in "ChatGPT, LLMs and Deep Learning with Python", which is written by Panagiotis Symeonidis and presents the most important algorithms of the last 20 years from the main scientific fields of Deep Learning, Transfer learning, Transformers, Information Retrieval, Machine Learning, Data Mining and Artificial Intelligence using as case study the Recommender Systems domain. The program is addressed to professionals in the field of Large Language Models and Deep Learning. It additionally provides developers with a comprehensive overview of the fundamental concepts of AI Systems, and presents all new methods and techniques through realistic and modern examples using the Python programming language. Indicative material of the training program (e.g., the book, the slides, the python programming exercises, etc.) can be found :
http://www.panagiotissymeonidis.com/llmbook/index.php

Learning outcomes:

Upon completion of the programme the trainee will be able to:
1. Understand and implement Deep Learning algorithms
2. Understand and implement Transformers and Large Language Models
3. Understand and implement Intelligent Recommendation Systems
4. Understand and implement Information Retrieval algorithms
5. Understand and implement Machine Learning algorithms
6. Understand and implement Data Mining algorithms
7. Understand and implement Artificial Intelligence algorithms

Implementation Methology:

[Online Distance Learning]
Blended Learning (Modern Distance Learning)

Evaluation Methology:

The evaluation will be with the delivery of the python code for 2 programming exercises

Minimum requirements of the candidates:

High School Certificate

Program Description

Subject & purpose of the training program:

Learning outcomes:

[Online Distance Learning]
Blended Learning (Modern Distance Learning)

Evaluation Methology:

The evaluation will be with the delivery of the python code for 2 programming exercises

Minimum requirements of the candidates:

High School Certificate

Program outline

Teaching Unit 1: Tranformers, Large Language Models, and ChatGPT

Description	1.1 Transformers 1.2 Attention and Seif-Attention 1.3 Multi-head Attention 1.4 Positional Encoding 1.5 Large Language Models and ChatGPT 1.5.1 Transformer layers 1.5.2. Positional Encoding 1.6 Encoder Transformers 1.7 Decoder Transformers 1.8 Encoder and Decoder Transformers 1.9 Multi-model Transformers and Google's Med Gemini 1.10 Chapter Questions
Start date	02/02/2025
End date	04/02/2025
Hours	10.00
Academic Responsibility	PANAGIOTIS SYMEONIDIS
Sessions	1 Transformers, Large Language Models, and ChatGPT

Teaching Unit 2: Large Language Models and Recommendation Algorithms

Description	Objectives, Types of Large Language Models, ChatGPT, Gemini, Llama and other Hybrid Systems, Types of Data Mining/Machine Learning/Transfer/Information Retrieval/Generative Adversarial Network Algorithms, large language models (LLMs) and large multi-modal models (LMMs), Audio transformer, Word transformer, Video transformer, Image transformer. 2.1 The Vector Space Model, bag of words and Tokenization Vector Space Model (TF-IDF), 2.2 Word Embedding, Word2Vec and skip-grams approaches 2.3 Recommendations based on Content Data 2.4 Decision Tree Classifier Decision Tree Classifier, Bayesian Classifier 2.5 Feature Selection Feature Selection, Gini Index, Entropy, X2-statistic 2.6 Naive Bayes Classifier 2.7 Python Exercise: Recommend Friends based on the CBF Algorithm
Start date	05/02/2025
End date	07/02/2025
Hours	10.00
Academic Responsibility	PANAGIOTIS SYMEONIDIS
Sessions	1 Large Language Models and Recommendation Algorithms

Teaching Unit 3: Collaborative Filtering Algorithms

Description	3.1 Introduction in CF algorithms Introduction in Information Retrieval and Filtering Factors and Challenges in Collaborative Filtering (CF), 3.2 Factors Affecting the CF Process First Stage Factors, Second Stage Factors, Third Stage Factors 3.3 Comparative Performance of CF algorithms Comparative Performance Study of CF Algorithms 3.4 Chapter Questions 3.5 Python Exercise: Constructing a User-User Similarity Matrix Implementation of the User-User Similarity Matrix in Python
Start date	08/02/2025
End date	10/02/2025
Hours	10.00
Academic Responsibility	PANAGIOTIS SYMEONIDIS
Sessions	1 Collaborative Filtering Algorithms

Teaching Unit 4: Context-aware AI Systems

Description	Sequence-aware AI systems, Sequence to Sequence Transformers Encoder-Decoder Transformers 4.1 Time-aware AI Systems 4.2 Location-aware AI Systems 4.3 Recommendation systems for LBSNs LBSNs = Location-based Social Networks 4.4 Types of Recommendations in LBSNs Location-aware Recommendation, Itinerary Recommendation, Hybrid Recommenders, Weighted Hybrid methods, Ensemble-based methods. 4.5 Hybrid and Ensemble Methods 4.6 Types of Hybrid Systems 4.7 Python Exercise: Recommend Friends based on a Hybrid Algorithm
Start date	10/02/2025
End date	12/02/2025
Hours	10.00
Academic Responsibility	PANAGIOTIS SYMEONIDIS
Sessions	1 Context-aware AI Systems

Teaching Unit 5: Dimensionality Reduction and Matrix Decomposition Algorithms

Description	Dimensionality Reduction, Matrix Decomposition, Principal Components Analysis, Singular Value Decomposition, UV-decomposition, Alternating Least Squares Method (wALS), Tensor decomposition, Bayesian Pairwise Ranking. 5.1 Eigenvalue Decomposition 5.2 Singular Value Decomposition 5.3 From SVD to UV-decomposition 5.4 Tensor Decomposition 5.5 From SVD to UV-decomposition 5.6 Tensor Decomposition 5.7 Tucker Decomposition 5.8 Python Exercise: Apply UV-decomposition to a User-Item Rating Matrix
Start date	12/02/2025
End date	14/02/2025
Hours	10.00
Academic Responsibility	PANAGIOTIS SYMEONIDIS
Sessions	1 Dimensionality Reduction and Matrix Decomposition Algorithms

Teaching Unit 6: Deep Neural Networks and Genetic Algorithms

Description	Multi-layer Perceptrons, Recurrent Neural Networks, Convolution Neural Networks, Graph Neural Networks, Feed Forward Propagation and Back-propagation algorithm. 6.1 Perceptron's structure 6.2 Single-layer Perceptron 6.3 Multi-layer Perceptron 6.4 Convolutional Neural Networks 6.5 Recurrent Neural Networks 6.6 Genetic Algorithms 6.7 Genetic Operations 6.8 Python Exercise: Use a Neural Network to Factorise a User-Item Rating Matrix
Start date	15/02/2025
End date	17/02/2025
Hours	10.00
Academic Responsibility	PANAGIOTIS SYMEONIDIS
Sessions	1 Deep Neural Networks and Genetic Algorithms

Teaching Unit 7: Deep Reinforcement Learning

Description	Markov Chain Model, Tabular Q-learning, Reinforcement Learning, A2C (Advantage Actor-Critic), Double Deep Q-Network, Genetic Algorithms 7.1 Q-learning Algorithm 7.2 Step-by-Step Execusion of the Q-learning Algorithm 7.3 Deep Reinforcement Learning 7.4 Deep Q-Network with Experience Replay Algorithm 7.5 Advantage Actor Critic Algorithm 7.6 Python Exercise 1: Implement the Tabular Q-learning Algorithm 7.7 Python Exercise 2: Implement the DQN Algorithm
Start date	17/02/2025
End date	19/02/2025
Hours	10.00
Academic Responsibility	PANAGIOTIS SYMEONIDIS
Sessions	1 Deep Reinforcement Learning

Teaching Unit 8: Deep Graph Neural Networks

Description	Graph Local-based Graph Similarity algorithms, Common Friends, Triangles, Connected Components, Global-based Graph Similarity algorithms, Random Walk with Restart, SimRank algorith, PathSim algorithm, Graph attention networks, Edge embeddings, Node embedding 8.1 Graphs Fundamentals 8.2 Local-based Similarity Measures 8.3 Global-based Similarity Measures 8.4 Knowledge Graphs 8.5 Graph Convolutional Networks 8.6 Python Exercise: Graph-based Recommendations for an Online Newspaper
Start date	20/02/2025
End date	22/02/2025
Hours	10.00
Academic Responsibility	PANAGIOTIS SYMEONIDIS
Sessions	1 Deep Graph Neural Networks

Teaching Unit 9: Evaluation metrics of AI systems

Description	Evaluation Metrics, Contingency Table, Precision, Recall, and NDCG, ROC curves, MAE, RMSE for Rating Prediction. 9.1 Introduction to Models' Evaluation 9.2 MAE and RMSE 9.3 Precision, Recall and F1 metric 9.4 ROC curve and AUC metric 9.5 Normalized Discounted Cumulative Gain 9.6 Beyond Accuracy Metrics 9.7 Python Exercise: Build an Evaluation Framework
Start date	23/02/2025
End date	25/02/2025
Hours	10.00
Academic Responsibility	PANAGIOTIS SYMEONIDIS
Sessions	1 Evaluation metrics of AI systems

Teaching Unit 10: New Trends in LLMs and Recommender Systems

Description	From LLMs to LMMs (multi-modal), Med Gemini, large language models (LLMs) and large multi-modal models (LMMs), Audio transformer, Word transformer, Video transformer, Image transformer. ChatGPT for Group recommenders 10.1 Group Recommender Systems 10.2 Ethics in Intelligent Systems 10.3 Fairness, Accountability, Censorship 10.4 Privacy in Intelligent Systems 10.5 Systems' Architecture for Privacy 10.6 Algorithmic Techniques for Privacy Protection 10.7 Legal Framework for Privacy Protection Privacy-aware recommenders, Types of Attacks, Privacy Protection Methods, Fairness, Accountability, Examples of non-biased recommender systems, Fair PageRank-based Random walks.
Start date	25/02/2025
End date	28/02/2025
Hours	10.00
Academic Responsibility	PANAGIOTIS SYMEONIDIS
Sessions	1 New Trends in LLMs and Recommender Systems

Contact:

Contact Details: PANAGIOTIS SYMEONIDIS [psymeon(at)aegean.gr - Phone 00306973882408]

WEBSITE: http://www.panagiotissymeonidis.com/llmbook/index.php

Discounts - Terms of use:

Discount Policy
No discounts are offered on the current training program

Registration
The application for registration is a Responsible Declaration and therefore the completed data must be true and accurate so that the issued Training Certificates, Certificates and other documents are valid. For any discrepancy that arises, CE-LLC is not obliged to reissue the above documents while it reserves the right to recall it.

Validity of submitted supporting documents
In order to document the validity of the discounts selected in the registration form, the trainee must send the relevant documents to the secretariat of the respective training program before the start of the selected program. According to article 28 of Law 4305/2014, the CE-LLC may carry out an ex officio check of the authenticity of the documents submitted by the trainee. In case of non-timely dispatch or non-confirmation of the validity of the submitted documents, the CE-LLC reserves the right to revoke the granted discount and to claim any resulting financial difference as well as to cancel any issue of any Certificate and/or Certificate of Training.

Visitor’s/user’s obligations
The user of the website of the educational platforms and the videoconferencing platforms must, on the one hand, comply with the rules and provisions of Greek, European and International Law and the relevant legislation governing telecommunications, and on the other hand, refrain from any illegal and abusive use of the website's content and services. It is expressly specified that personal access codes may not be given to any third party. Also, every visitor / user of the website must behave politely and discreetly during the use of the above, while the adoption of practices of unfair competition or others that are contrary to the commonly accepted rules of behavior of Internet users (Netiquette) is prohibited.
Any damage caused to the website or/and the platforms of modern/ asynchronous education or the Network in general resulting from the misuse or improper use of the relevant services by the user/visitor is solely his/her own responsibility, and the University of the Aegean may delete him/her from its educational community without any financial compensation and reimbursement of tuition fees. In addition, any moral or reputational damage caused by the user's failure to use proper and discrete behavior in his/her communication with the individual teaching, technical and administrative support teams is the sole responsibility of the user, and the University of Aegean may remove him/her from the educational community without any financial compensation and refund of fees. For any comments, remarks or/and suggestions, the trainees should be addressed exclusively to the secretariats of the individual training programs or to the secretariat of CE-LLC.

Participation in courses and Issuance of Certificate
- There is no restriction of gender, race, nationality, religion or other discrimination in participation in the training programs of CE-LLC.
- The acceptance and participation of the trainees re-adjust their full understanding of the program and the respective Program Guide
- The CE-LLC, reserves the right to adjust the schedule of the courses, the instructors as well as to postpone or cancel the availability of the Programs by informing the registered participants.
- In order to serve them better, the trainees should mention while contacting with the CE-LLC, their full name, the exact title of the program they are participating in, the Academic Coordinator and the start date of the program.
- If there are outstanding financial issues, the CE-LLC reserves the right to temporarily suspend or/and deactivate the access of the trainees to the e-learning platforms of modern and asynchronous learning.
- The award of the Certificate of Training requires the participation of the trainees in the assigned evaluation activities and on the basis of the evaluation criteria set by the respective training program.
- If there are outstanding financial issues, no Training Certificate will be awarded for any of the individual teaching modules of a Program, nor for the whole Program.

- A prerequisite for the granting of certificates before the completion of a training program is the payment of at least part of the tuition corresponding to the Educational Modules that it is confirmed that the trainee attended.

Technical Conditions
To participate in the online training, a modern computer and internet access are required at least.
Specifically, for optimal and uninterrupted access and work in the Asynchronous and especially the Modern e-learning System, it is recommended that the interested parties have the following:
1. Modern computer with multimedia capabilities (microphone, speakers or headphones, camera).
2. ADSL connection for Internet access, with a minimum required Internet connection speed of 4 Mbps.
However, additional technical requirements may be required by the individual training programs, if deemed appropriate for their smooth running.

Copyright
Any material, trademark or other intellectual property of this website is the property of the University of the Aegean and is subject to the relevant legislation on the protection of intellectual property, with the exception of the expressly recognized rights of third parties. Its reproduction, modification/copying/rental/borrowing/broadcasting and transmission without permission is prohibited. Exceptionally, the individual storage and copying of parts of the content on a simple personal computer for strictly personal use is permitted, without any intention of commercial or other exploitation and always provided that the source of the content is indicated, without this in any way implying the granting of intellectual property rights.

Prioritisation of Programs
In the context of promoting a program, the CE-LLC may inform -only in an advisory capacity- the interested public about the possibility of using a training certificate in terms of its possible awarding, based on the institutional framework in force at the time of the program's launch. However, the classification of training as a qualification depends on the criteria of the calls for proposals as defined by the bodies issuing them at the time of their issue, and the interested parties should contact these bodies for clarification and are solely responsible for investigating classification issues.

Cancellation of Programs - Tuition Refunds
The training programs will only be implemented if the required number of trainees has been secured.
A refund of tuition fees is possible only in exceptional cases and only if the applicant informs the secretariat in writing before the start of the program that he/she does not wish to attend, stating in detail the reasons for discontinuing the training. The written request is examined by the Scientific Coordinator of each training program and he/she decides on it. In such cases and in order not to be burdened with possible bank charges or administration costs of the Special Account for Research Funding, the student may apply for a credit of his/her tuition fees for attending another Program of the same training action, i.e. a Program with the same Scientific Coordinator. The amount credited, if not used up, is not refundable, but remains at the participant's disposal for future use.
In cases of tuition refund decisions, if the CE-LLC is responsible for the refund, the financial difference due is refunded without any additional charge.
In cases of non-culpability of the University of the Aegean, the applicant is responsible for any financial difference due, bearing any bank charges and an additional 10% of the administration costs of the Special Account for Research Funds.

Program evaluation
Participation in the Program Evaluation: The completion of the anonymous Training Program Evaluation Questionnaire by the trainees aims to ensure the quality of the educational services provided, in accordance with the provisions of Article 115 of Law No. 4957/2022 and the Regulation of the CE-LLC of the University of the Aegean. The questionnaire is posted on the asynchronous tele-education platform during the last month of the program implementation, with the activation of the last evaluation activity of the program together. The completion of the questionnaire is optional and is part of the academic activities of the trainees.

Extension of academic obligations
Any extension in the academic obligations of the trainees as well as the procedure for issuing the training certificates for the trainee(s) who have successfully completed the program may not exceed six (6) months from the date of completion of the program, based on the approval decision. In exceptional cases, following a well-founded request by a trainee and a reasoned decision by the Council of the CE-LLC, the extension provided for in the previous subparagraph may be granted for a period of up to one (1) year from the date of completion of the program, based on the approval decision.