Applied AI Study Group #4 Report

Offering an interactive platform to explore artificial intelligence through hands-on coding for the fourth time in inzva, another batch of Applied AI Study Group has come to an end after 4 weeks full of fun and productive learning experience. Under the guidance of talented AI enthusiasts from our community -Ahmet Melek, Halil Eralp Kocaş, Onur Boyar, and Uğur Ali Kaplan, -, 25 graduates from 17 different universities acquired knowledge about various topics including, but not limited to, computer vision, NLP, and anomaly detection along with developing related skills in machine learning.

w4_ended_tw.jpeg

Following the increasing trend in virtual activities due to the pandemic, though we miss having packed coding sessions in our sanctuary in Beykoz Kundura, due to the pandemic we continue holding our sessions online without losing anything of the program’s dynamic character while providing the participants the opportunity to attend the meetings from wherever they want.

During the program period, participants met with our experienced program guides every Saturday to discuss various topics on artificial intelligence. They also received three assignments, to practice their recently obtained knowledge and skills, and these assignments were followed by lab sessions led by our guides where their questions and comments as to the assignments were discussed. The entire content of our community-driven program is open-sourced and uploaded on our Github repository and all the session recordings are available on our YouTube Channel. Do not hesitate to follow the course from the videos and solve the notebooks to practice your skills.

 
 

Check below for the topics discussed each week to learn more about the final program of 2020.

WEEK 1: COMPUTER VISION led by Halil Eralp Kocaş

The first week of the program concentrated on one of the core problems in machine learning which is classification. While doing so, various deep learning frameworks including TensorFlow, Keras, and PyTorch were used. This first part also dealt with the topics of localizing and detecting the objects as well as providing participants a hands-on experience in computer vision.

We prepared two homework for participants to make use of their new knowledge; the first one, prepared by Ahmet Melek, focused on data preprocessing skills and model building skills on computer vision. The second one, prepared by Halil Eralp Kocaş, was a fun challenge for participants that do not only evaluate their skills but also give room for some competition. 

WEEK 2: ANOMALY DETECTION led by Onur Boyar

The second week started with the anomaly detection models and their data preprocessing. Subsequently, a wide range of methods used in anomaly detection was discussed, namely LightGBM, XGBoost, Isolation Forest, Autoencoders, and more. This week’s homework challenged the participants to deep dive into concepts of Gradient Boosting trees and different hyperparameter optimization techniques.

WEEK 3: NATURAL LANGUAGE PROCESSING led by Ahmet Melek

This week took a deep dive into NLP challenges including language modeling, neural machine translation, summarization, and named entity recognition. Moreover, the participants learned about preprocessing and data formats, frameworks, experimentation, remote development, deployment, and staying current with the state-of-the-art research.

The homework for the week was not only designed in a way that makes the participants with existing scraping tools and using them to scrape an NLP dataset from the web but also in a way that familiarizes them with the state of the art natural language generation architecture and some niche natural language processing tools such as the attention mechanism.


WEEK 4: EXPERIMENT TRACKING, DATA PIPELINES, DEPLOYMENT, END TO END SYSTEMS led by Uğur Ali Kaplan

The last part of the program took the participants through a journey in which they were introduced to more structured methods in machine learning. Throughout this journey, an exploratory data analysis was conducted and some predictive models were analyzed as well as developing models and performing hyperparameter tuning with MLflow. Further, the group discussed the ML models in the respects of deployment and interoperability.

We are happy to say that according to all feedbacks we got from our participants, we will be increasing the number of sessions held within this program to at least 6 weeks for the next batch, which will take off around mid-Summer!


A  BEV Foundation project inzva is a non-profit hacker community organizing study and project groups as well as camps in the fields of AI and Algorithm; and gathering CS students, academics and professionals in Turkey.

Follow us on our social media accounts to have the most recent news about our upcoming events and programs!