Why Training data and models in Intro to Computing? - Purpose & Use Cases

The Big Idea

What if your computer could learn from examples just like you do, without you telling it every single rule?

The Scenario

Imagine you want to teach a computer to recognize pictures of cats and dogs by showing it thousands of photos and telling it which is which by hand.

You write down notes for each photo, then try to guess the rules yourself to separate cats from dogs.

The Problem

This manual way is slow and tiring because you must look at every photo and write down details.

It's easy to make mistakes or miss important patterns that are hard to see with just your eyes.

Also, as the number of photos grows, it becomes impossible to keep up.

The Solution

Training data and models let the computer learn patterns by itself from many examples.

You give the computer labeled photos (training data), and it builds a model that can guess new photos automatically.

This saves time, reduces errors, and can find hidden patterns humans might miss.

Before vs After

✗ Before

for photo in photos:
    if 'pointy ears' in photo and 'whiskers' in photo:
        label = 'cat'
    else:
        label = 'dog'

✓ After

model = train_model(training_photos, labels)
predictions = model.predict(new_photos)

What It Enables

It enables computers to learn from examples and make smart decisions without explicit instructions for every case.

Real Life Example

Streaming services use training data and models to recommend movies you might like based on what you watched before.

Key Takeaways

Manual sorting of data is slow and error-prone.

Training data teaches computers by example.

Models learn patterns to make predictions automatically.

Practice

1. What is the main purpose of training data in machine learning?

easy

A. To teach the model patterns and relationships in data

B. To test the model's accuracy after training

C. To store the final predictions of the model

D. To clean and prepare data for analysis

Solution

1. Step 1: Understand the role of training data

   Training data is used to show the model examples so it can learn patterns.

2. Step 2: Differentiate training from testing

   Testing data checks model accuracy, but training data teaches the model.

3. Final Answer:

   To teach the model patterns and relationships in data -> Option A

4. Quick Check:

   Training data = Teach model [OK]

Hint: Training data is what the model learns from [OK]

Common Mistakes:

• Confusing training data with testing data
• Thinking training data stores predictions
• Assuming training data is for cleaning only

2. Which of the following is the correct way to split data into training and testing sets in Python using scikit-learn?

easy

A. test_train_split(data, size=0.2)

B. train_test_split(data, test_size=0.2)

C. train_test(data, split=0.2)

D. split_train_test(data, 0.2)

Solution

1. Step 1: Recall the scikit-learn function name

   The correct function to split data is train_test_split.

2. Step 2: Check the parameter syntax

   The parameter test_size=0.2 correctly sets 20% data for testing.

3. Final Answer:

   train_test_split(data, test_size=0.2) -> Option B

4. Quick Check:

   Correct function and parameter = train_test_split(data, test_size=0.2) [OK]

Hint: Remember scikit-learn's function is train_test_split [OK]

Common Mistakes:

• Using wrong function names
• Mixing parameter names
• Incorrect order of arguments

3. Given the following Python code, what will be the output?

from sklearn.linear_model import LinearRegression
import numpy as np

X = np.array([[1], [2], [3], [4]])
y = np.array([2, 4, 6, 8])
model = LinearRegression()
model.fit(X, y)
pred = model.predict([[5]])
print(round(pred[0], 1))

medium

A. 9.0

B. 12.0

C. 8.0

D. 10.0

Solution

1. Step 1: Understand the training data relationship

   X values are 1 to 4, y values are double X (2,4,6,8), so model learns y = 2 * X.

2. Step 2: Predict for X=5 using the model

   Prediction = 2 * 5 = 10.0, rounding to 1 decimal is 10.0.

3. Final Answer:

   10.0 -> Option D

4. Quick Check:

   Prediction for 5 = 10.0 [OK]

Hint: Model learns y=2x, predict for 5 gives 10 [OK]

Common Mistakes:

• Ignoring rounding in output
• Misunderstanding model relationship
• Using wrong input shape for predict

4. Identify the error in this code snippet for training a model:

from sklearn.linear_model import LogisticRegression
X_train = [[0, 1], [1, 0]]
y_train = [0, 1]
model = LogisticRegression()
model.fit(y_train, X_train)

medium

A. The order of arguments in model.fit() is reversed

B. LogisticRegression cannot be used with this data

C. X_train should be a numpy array, not a list

D. y_train should contain strings, not integers

Solution

1. Step 1: Check the model.fit() argument order

   fit() expects features first (X_train), then labels (y_train).

2. Step 2: Identify the mistake in the code

   Code passes y_train first, then X_train, which is incorrect and causes error.

3. Final Answer:

   The order of arguments in model.fit() is reversed -> Option A

4. Quick Check:

   fit(X_train, y_train) correct order [OK]

Hint: fit() needs features first, labels second [OK]

Common Mistakes:

• Swapping X and y in fit()
• Assuming data type causes error here
• Thinking LogisticRegression needs string labels

5. You have a dataset with 1000 samples and 10 features. You want to train a model but avoid overfitting. Which approach helps most during training?

hard

A. Use all data for training without splitting

B. Train only on the first 100 samples

C. Split data into training and validation sets

D. Ignore feature scaling and train directly

Solution

1. Step 1: Understand overfitting and validation

   Overfitting happens when model learns training data too well and fails on new data.

2. Step 2: Use validation set to check model generalization

   Splitting data into training and validation helps detect and reduce overfitting.

3. Final Answer:

   Split data into training and validation sets -> Option C

4. Quick Check:

   Validation split prevents overfitting [OK]

Hint: Validation set helps check model generalization [OK]

Common Mistakes:

• Training on all data hides overfitting
• Using too few samples reduces learning
• Ignoring feature scaling can hurt model