🧠 Deep Learning Foundations

I started the week by strengthening my understanding of core deep learning concepts:

• What is deep learning?

• What is a neuron?

• Types of neural networks

• Deep learning architectures

This helped me build a solid base before moving into implementation.

🔢 Building My First Deep Learning Model (MNIST)

One of the most exciting parts of this week was implementing a deep learning model using the MNIST dataset (handwritten digit recognition).

Each image in the dataset has:

• Size: 28 × 28 pixels

• Total features: 784

Key things I learned:

✅ Data Preprocessing

Before training a model, data needs to be normalized.
This helps:

• Stabilize training

• Improve performance

• Ensure faster convergence

🧱 Understanding Layers

I explored:

• Flatten Layer → Converts 2D image data into a 1D vector

• Dense Layer → Fully connected layer used for learning patterns

I built a sequential model, where layers are stacked one after another.

⚙️ Training the Model

While training, I learned:

• What are epochs and batch size

• How loss functions work

• Role of optimizers

• How models improve over iterations

📊 Model Evaluation

After training, I compared training and testing results to understand:

• Overfitting

• Underfitting

• Best-fit scenarios

This gave me a practical understanding of model performance.

🔀 Exploring More Complex Architectures

After building a simple model, I experimented with a more complex one:

• Used a non-sequential (functional) model

• Added multiple hidden layers

• Increased the number of neurons

I also learned how to merge layers using:

• Concatenate

This helped me understand how flexible model architectures can be designed.

🤖 Introduction to Generative AI & Attention

This week also marked my entry into Generative AI concepts.

I learned about:

• Attention mechanisms

• Why they are important in modern AI

🔍 Understanding Attention

I explored:

• Query (Q), Key (K), Value (V)

• How words are converted into numerical representations

• How models understand relationships between words

🧩 Single-Head vs Multi-Head Attention

• Single-head attention → Focuses on one representation

• Multi-head attention → Captures multiple perspectives

👉 Multi-head attention is more powerful because it understands context better.

🔹 Getting Started with Transformers

Towards the end of the week, I began exploring transformers.

This was more of an introduction rather than a deep dive.

📈 Reflection

This week felt like a shift from just learning concepts to actually understanding how things work.

It’s getting a bit challenging, but also more interesting.

🔜 What’s Next?

Next, I plan to:

• Learn transformers in more detail

• Explore more about Generative AI

✨ Final Thought

I’m still a beginner, but this week helped me understand that learning AI is all about practicing and building step by step.