Why agents make autonomous decisions in Prompt Engineering / GenAI - Model Pipeline Impact

Model Pipeline - Why agents make autonomous decisions

This pipeline shows how an autonomous agent learns to make decisions on its own by observing data, learning patterns, and improving its choices over time.

Data Flow - 6 Stages

1Data Collection

1000 rows x 6 columns→Gather sensor readings and environment states→1000 rows x 6 columns

Row example: [temperature=22, light=300, obstacle=0, speed=5, direction=90, reward=1]

↓

2Preprocessing

1000 rows x 6 columns→Normalize sensor values and encode categorical data→1000 rows x 6 columns

Normalized speed from 5 to 0.5, direction encoded as angle in radians

↓

3Feature Engineering

1000 rows x 6 columns→Create new features like distance to obstacle and speed change→1000 rows x 8 columns

Added features: distance_to_obstacle=10, speed_change=0.1

↓

4Model Training

800 rows x 8 columns→Train decision-making model on training data→Trained model

Model learns to predict best action given sensor inputs

↓

5Validation

200 rows x 8 columns→Test model on unseen data to check accuracy→Validation accuracy and loss metrics

Accuracy=0.85, Loss=0.35

↓

6Autonomous Decision Making

New sensor data (1 row x 8 columns)→Model predicts best action to take→Action decision (e.g., move forward, turn left)

Input: sensor readings; Output: action=turn left

Training Trace - Epoch by Epoch

Loss
0.8 |****
0.7 |*** 
0.6 |**  
0.5 |*   
0.4 |*   
0.3 |    
     1 2 3 4 5 Epochs

Epoch	Loss ↓	Accuracy ↑	Observation
1	0.75	0.5	Model starts learning, accuracy at chance level
2	0.6	0.65	Loss decreases, accuracy improves
3	0.48	0.75	Model learns important patterns
4	0.4	0.8	Better decision making emerging
5	0.35	0.85	Model converges with good accuracy

Prediction Trace - 4 Layers

Layer 1: Input sensor data

Layer 2: Model hidden layers

Layer 3: Output layer

Layer 4: Decision

Model Quiz - 3 Questions

What happens to the data shape after feature engineering?

A It decreases from 6 to 4 columns

B It stays the same at 6 columns

C It increases from 6 to 8 columns

D It changes from rows to columns

Key Insight

Autonomous agents learn to make decisions by transforming raw sensor data into meaningful features, training a model to predict the best actions, and improving accuracy over time through repeated learning.