data310

«««< HEAD Wednesday’s Response

  1. img.png img_1.png img_2.png
    • predictions[3] arrays ([3.6671641e-09, 1.0000000e+00, 8.4066903e-12, 2.8593914e-09, 1.2264663e-09, 4.9193018e-22, 1.2822771e-09, 4.5023612e-27, 1.0708230e-16, 2.3360760e-15], dtype=float32)
    • The numbers in the array represent the confidence in its prediction that the image corresponds to the correct label.
    • Softmax turned the output of the code “probability_model = tf.keras.Sequential([model, tf.keras.layers.Softmax()])” into probability distributions so they can be interpreted by the same scale.
    • Argmax was used to return the number of the predictions array of the image from the array with the highest probability of being correct, from the code ‘np.argmax(predictions[2500])’.
    • Yes the argmax from the predictions array matched the label from the test_labels dataset using np.argmax(predictions[1050]) and test_labels[1050], they both returned 4.

    • img_5.png

    • img_6.png

    • img_7.png
    • No the predicted value did not match the test label.
    • We did not sue Softmax when determining if the test label matched the predicted value of the trained model because in the line ‘predictions_single = probability_model.predict(img)’ we use ‘probability_model’ which we already defined to have the Softmax function in it back in the ‘Make Predictions’ section.
    • The accuracy of the training dataset was 0.9980
    • The accuracy of the testing dataset was 0.9774 img_9.png

    img_10.png

    • The training data set was more accurate, I think this is because our dataset was small and we only had ten classifications and the numbers look very different from on another making it easier to train a model.

      Wednesday’s Response

  6. img.png img_1.png img_2.png
    • predictions[3] array ([3.6671641e-09, 1.0000000e+00, 8.4066903e-12, 2.8593914e-09, 1.2264663e-09, 4.9193018e-22, 1.2822771e-09, 4.5023612e-27, 1.0708230e-16, 2.3360760e-15], dtype=float32)
    • The numbers in the array represent the confidence in its prediction that the image corresponds to the correct label.
    • Softmax turned the output of the code “probability_model = tf.keras.Sequential([model, tf.keras.layers.Softmax()])” into probability distributions so they can be interpreted by the same scale.
    • Argmax was used to return the number of the predictions array of the image from the array with the highest probability of being correct, from the code ‘np.argmax(predictions[2500])’.
    • Yes the argmax from the predictions array matched the label from the test_labels dataset using np.argmax(predictions[1050]) and test_labels[1050], they both returned 4.

    • img_5.png

    • img_6.png

    • img_7.png
    • No the predicted value did not match the test label.
    • We did not sue Softmax when determining if the test label matched the predicted value of the trained model because in the line ‘predictions_single = probability_model.predict(img)’ we use ‘probability_model’ which we already defined to have the Softmax function in it back in the ‘Make Predictions’ section.
    • The accuracy of the training dataset was 0.9980
    • The accuracy of the testing dataset was 0.9774 img_9.png

    img_10.png

    • The training data set was more accurate, I think this is because our dataset was small and we only had ten classifications and the numbers look very different from on another making it easier to train a model.

      origin/main