O događaju
Radionica "Duboko učenje - osnovni alat umjetne inteligencije"
Workshop „Deep Learning - Basic Tool in Artificial Intelligence “
Sadržaj/Content
Radionica pokriva osnovne ideje, pristupe, tehnike i primjene neuronskih mreža (NM), dubokog učenja (DU) i umjetne inteligencije (UI) i namijenjena je svima zainteresiranima u primjeni ovih alata u njihovom području interesa. Bit će predstavljene ključne tehnike u upotrebi danas – povratno vraćanje pogreške, neuronska mreža s jednim slojem i njeno produžavanje na više slojeva, poznato danas kao duboko učenje. Radionica će kombinirati interaktivna predavanja i primjere sa demonstracijama u Pythonu sa ciljem usavršavanja ideja i koncepata. Na taj će se način polaznici upoznati sa alatima koje će poslije uspješno primjenjivati u svojim projektima i zadacima. Radionica neće biti fokusirana na specifične aplikacije već će predstaviti širok uvod u uzbudljive ideje i pristupe u umjetnoj inteligenciji danas.
Sve aktivnosti će se provesti online – preko Microsoft Teams platforme.
Jezik komunikacije tijekom radionice: Engleski jezik
The workshop will cover basic ideas, approaches, techniques, and applications of Neural Networks (NN), Deep Learning (DL), and Artificial Intelligence (AI) and it is aimed at all interested in using these tools in their field of interest. It will cover the core techniques used today and in particular, it plans to introduce error backpropagation algorithm (EBP), single layer NNs and their extensions to multilayers learning structures dubbed DL. The course will combine interactive lectures, exercises and Python code demonstrations aiming at mastering the concepts and get acquainted with the tools which would later enable all the participants to use them in their applications and projects. The workshop will not be focused on specific applications but rather give a broad introduction to the exciting new ideas and approaches in AI. The program includes hands-on sessions, demonstrating practical aspects.
All activities will be online – on the Microsoft Teams platform.
We are looking forward to seeing you participate in this workshop.
Predavači/Lecturers
Professor Tomasz Arodz, prof.dr.sc., Virginia Commonwealth University, Richmond, VA
Professor Vojislav Kecman, prof.dr.sc., VsiTe, College for Information Technologies, Zagreb
Program
Dan 1. Utorak, 23. Ožujaka 2021. / Day 1. Tuesday, 23 March 2021
Prijepodnevni dio, 9 – 12 AM, Prof. Kecman / Morning session
Basics of Machine Learning
Supervised vs Unsupervised Learning
Feed-Forward Neural Networks
Error Back Propagation
Bias and Variance in Neural Networks
Poslijepodnevni dio, 2 – 5 PM, Prof. Arodz / Afternoon session
Automated Differentiation (AD)
PyTorch example of Automated Differentiation
Problems in training Deep Networks and how to overcome them.
Convolutional Neural Networks (CNNs)
Residual Networks (ResNets)
Self-attention-based Networks (Transformers)
Dan 2. Srijeda, 24. Ožujaka 2021. / Day 2. Wednesday, 24 March 2021
Prijepodnevni dio, 9 – 12 AM, / Morning session
Exercise Session
Poslijepodnevni dio, 2 – 5 PM, Prof. Arodz / Afternoon session
Building deep networks from modules in PyTorch
PyTorch example of Convolutional Deep Network for image recognition
PyTorch example of an Attention-based Deep Network for language tasks
TensorFlow as an alternative to PyTorch
Molimo da uzmete u obzir da su minimalni preduvjeti za praćenje radionice osnovno znanje Python programskog jezika, osnovno poznavanje matričnog računa i analize funkcije jedne varijable.
Please consider that the required prerequisites are: at least basic knowledge of Python, a basic understanding of matrix algebra, and single variable calculus.