Ciklum Speaker’s Corner – Neural network that can learn: creating from scratch (Ciklum, 2023)
Info
- Project type Public Speaking, Tech Presentation
- Date 21st of Sept 2023
- My Role Speaker
- Registered 448, from 14 countries
- Topics AI, Neural Networks, Learning
- Keywords Neuron, Synapse, Learning, Emergence, Weights, Activation function, Loss function, Gradient, Learning rate, Neural network parameters, Training, Testing, Epochs, Neural network architecture, Neural layers, Deep Learning
- Skills developed Presentation, Speaking
- Event URL Ciklum's website
- Event Promo video
-
Event recording (Youtube)
Description
This was the second time I conducted a live coding session in front of an international audience of AI professionals—developers, DevOps engineers, data scientists, and AI salespeople. I had the great opportunity to demonstrate the fundamental concepts of neural networks at the very heart of machine learning during a ~90-minute live demo session at Ciklum Speaker's Corner.
Preparing for such events is demanding, making it a challenging experience. I learned that writing the code and setting up the environment is not the complicated part—the real challenge lies in handling exceptions and unexpected scenarios during the demo.
I appreciate the opportunity provided by my colleague Iryna Rud and thank everyone for their feedback.
This session drew on the knowledge I've gained during my Ph.D. program in AI and while preparing for my AI Developer certification. During my certification preparation, I also worked on publishing a series of articles to better understand and organize the fundamentals in my mind.
The aim of this session was to demonstrate how to create a neural network from scratch using Java and Python. I also used Javascript to build illustrations and animations to highlight the "magic" behind machine learning.
This event posed several new challenges for me:
- Presenting to a global audience and to unknown attendees
- Engaging a diverse group with varying levels of expertise and skill sets
- Representing a strong brand like Ciklum in public
- Being broadcasted and recorded without the option for retakes
This was an experience that required intensive preparation and pushed me to overcome my limitations. I learned a lot and received valuable feedback.
I'm especially grateful to Ciklum's branding team for the invitation and opportunity. Thank you!
References
- Github repo
- Presentation slides
- Video recording (Ciklum's YouTube channel)
- My articles related to this topic
Tech stack
- Programming Languages: Java 17, Python 3.11, Javascript
- Java stack: Spring Boot, Spring Web, Thymeleaf, Lombok, org.json, Log4j
- Python stack: numpy, tensorflow, matplotlib, scikit-image
- Javascript stack: d3js
