Scribd
Upload
4 views5 pages

Perceptron Feature Analysis Results

The document presents results from a deep learning lab exercise comparing different feature sets in a Perceptron, highlighting accuracy, precision, recall, and F1-scores for various combinations. It also discusses hyperparameter tuning, identifying the best parameters for optimal model performance. Additionally, a summary of the best model configuration is provided, detailing its architecture, training efficiency, and classification metrics.

Uploaded by

khanna2004kartikey
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
4 views5 pages

Perceptron Feature Analysis Results

The document presents results from a deep learning lab exercise comparing different feature sets in a Perceptron, highlighting accuracy, precision, recall, and F1-scores for various combinations. It also discusses hyperparameter tuning, identifying the best parameters for optimal model performance. Additionally, a summary of the best model configuration is provided, detailing its architecture, training efficiency, and classification metrics.

Uploaded by

khanna2004kartikey
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd

Deep Learning Lab1

KARTIKEY KHANNA
23BAI1283

Exercise 1
Result comparison between different features in a Perceptron
Precision (Class Recall (Class F1-Score
Count Selected Features Accuracy TN FP FN TP
1) 1) (Class 1)

sepal length (cm), sepal


1 0.95 9 0 1 10 1.00 0.91 0.95
width (cm)

sepal length (cm), petal


2 1.00 5 0 0 15 1.00 1.00 1.00
length (cm)

sepal length (cm), petal


3 1.00 12 0 0 8 1.00 1.00 1.00
width (cm)

sepal width (cm), petal


4 1.00 7 0 0 13 1.00 1.00 1.00
length (cm)

sepal width (cm), petal


5 1.00 10 0 0 10 1.00 1.00 1.00
width (cm)

petal length (cm), petal


6 1.00 6 0 0 14 1.00 1.00 1.00
width (cm)
Model Interpretation
Model Weights (Feature Importance):
sepal length (cm): 0.0419
sepal width (cm): -0.0695
petal length (cm): 0.1061
petal width (cm): 0.1097
Model Bias (Intercept): [-0.3]
Exercise 1.2 Hyper tuning Parameters

Max_ite r = [20, 40, 80, 100, 200]


eta0 = [0.0001, 0.001, 0.01, 0.1, 1]
result: Best Parameters: {'eta0': 0.0001, 'max_iter': 20, 'penalty': None}
Best Accuracy: 1.0

MLP
Best Model Run Summary
Model Configuration

Arch: (100,50)
Alpha:0.01
Solver:adam
The best-performing model configuration utilized a two-layer architecture with hidden layer
sizes of 100 and 50 neurons, trained using the Adam optimizer with a learning rate of 0.01.
This setup achieved strong classification performance, with an accuracy of 96.49%, precision
of 98.57%, recall of 95.83%, and an F1-score of 97.18%. The model was also
computationally efficient, completing training in 1.68 seconds

Metric Malignant Benign Overall / Average

Precision 0.93 0.99 Accuracy: 0.96

Recall 0.98 0.96 Macro Avg: 0.96

F1-Score 0.95 0.97 Weighted Avg: 0.97

Support 42 72 114

You might also like