Let’s Learn How We Can “Make our Own AI Model and Make it Learn” (Part 1, Keywords) | by Furkan BAYHOCA | Apr, 2024

In last 2 or 3 years, the AI industry got so many advancements with the help of the “GPU”s and other factors. So many people say that it can make software engineers lose their jobs. I don’t think so. It wouldn’t make them lose their jobs, this can even help them evolve, make them get better at their jobs. In fact, This can help some people who knows how to develop AI to get better jobs and salaries.

Now, to get better jobs and salaries; You need to learn AI today, not tomorrow, today. If you say “It is too complex” don’t worry. It would get more simple as you learn. And these articles are here to help you. We need to inspect so many steps, but I will say it again, don’t worry. So, if we have relaxed, now let’s get started.

First thing first, we have to learn basic terms to understand each other.

1. Neural Network: Also referred as “AI Model”. This is the place where all the job is done. Contains so many layers.
2. Layer: This neural networks have so many layers that hold info. There are so many connections and combinations between these things. When we train the model, data goes between these layers and at the end, a trained model is being produced. (Types: Conv2d, Conv3d, LSTM, Dense…)
3. Activation Functions: We use them to give the neural network non-linear values or properties. (“softmax”, “sigmoid”, “tanh”, “ReLU”…)
4. Loss Functions: A way to exam our model. When you teach a student something, you need to make quizes so you can understand how good did he/she learn.It is same in the Deep Learning. (MSE, MAE, cross-entropy…)
5. Tensor: A mathematical shape to figure out 3D vectors. (Usually created with numpy)
6. Filtering: A process to test the quality of result.

These terms are enough for now. I think you can now partly imagine what will we do. If we need more terms in the future articles, I will give them with their explanations. That’s all for now, Bye!