What the basic components of AI are and how the data flows

[Chip (OP)]

What are the basic components of AI and show the data flow

Core components of AI (neural network model)

1. Input data
Raw information such as:
- images, text, audio, sensor readings 
Converted into numeric form (vectors/tensors)

2. Parameters (weights + biases)
- Weights: control strength of connections 
- Biases: shift outputs 
These are the “learned memory” of the model.

3. Layers (transformations)
Each layer computes:

Code: [Select]

output = activation(Wx + b)
They progressively transform raw input into useful representations.

4. Activation function
Adds non-linearity:
- ReLU, sigmoid, tanh 
Without it, the model becomes just a linear function.

5. Loss function
Measures error:
- difference between prediction and truth 
Outputs a single number representing “how wrong” the model is.

6. Optimiser
Updates weights to reduce error:
- gradient descent / Adam 
It performs learning.

---

Data flow (forward + learning loop)

Forward pass

Code: [Select]

Input
 ↓
Layer 1 (Wx + b + activation)
 ↓
Layer 2
 ↓
...
 ↓
Output (prediction)

Loss calculation

Code: [Select]

Prediction + True label → Loss → Error value

Backward pass (backpropagation)

Code: [Select]

Error
 ↓
Compute gradients (credit assignment)
 ↓
Propagate backwards through layers

Parameter update

Code: [Select]

Weights = Weights - learning_rate × gradient

---

Full AI training loop

Code: [Select]

Input → Forward pass → Prediction
                    ↓
                   Loss
                    ↓
            Backpropagation
                    ↓
           Update parameters
                    ↓
              Repeat many times

---

Key idea
AI is not rules or logic.
It is a feedback loop that tunes a large mathematical function until its outputs match data.

Generated by ChatGPT.