Lab 3 Hand In

Submit your results to Lab 3 by responding here. Your response should include the answers to the following:

1. What type of processing unit did you use (e.g. CPU, TPU, T4 GPU, etc.),
2. What was your random seed, and
3. What was the resulting accuracy for each of the 2 or 3 techniques that you tried?

And do be sure to include a link to your notebook!!

1. MLP CNN
2. 95

CNN: 0.9261
MLP: 8833

1. T4 GPU
2. Random seed: 33

Single layer perceptron: Perceptron Accuracy: 0.8433
A Multi-Layer Perceptron (MLP): MLP accuracy: 0.8823
CNN: CNN accuracy: 0.9302

1. T4 GPU
2. Random seed: 16
3. CNN accuracy: 0.9261
   Perceptron Accuracy: 0.8386

https://colab.research.google.com/drive/13a6NYtOSMQaOCYB2_fSEq_T3D1IIwOkO?usp=sharing

1. T4 GPU
2. Random Seed: 35
3. Perceptron: 84.35%, MLP: 87.98%, CNN: 92.54%

I used a T4 GPU with a random seed of 63.

I tried Perceptron, and CNN.

I got a Perceptron training time of 94.26 seconds
and a Perceptron Accuracy of 0.8434.

As well as a CNN training time of 268.42 seconds, with a
CNN accuracy of 0.9259

1. T4 GPU
2. 88
3. Single layer = .8423
   Multi-Layer = .8865
   CNN = .9275

Copy of Lab3NNForFashionMNIST.ipynb - Colab

1. T4 GPU
2. Random Seed: 98
3. Perceptron: 0.8459, MLP: 0.8825, CNN: 0.9252

Notebook

1. T4 GPU
2. Random Seed - 98
3. a. Perceptron Accuracy: .8452 (training time 73.52 sec)
   b. MLP Accuracy: .8857 (training time 214.63 sec)
   c. CNN Accuracy: .9230 (training time 195.06 sec)

1. T4 GPU
2. Random Seed: 17
3. Perceptron: 0.8448, MLP: 0.8828, CNN: 0.9244

T4 GPU
Random Seed: 41
Perception: 0.8352
MLP: 0.8799
CNN: 0.9261

Link:

1. T4 GPU
2. Random seed: 14
3. Single perceptron accuracy: 0.8433
   Multi-Layer Perceptron: 0.8828
   CNN accuracy: 0.9245
   Colab: Google Colab

1.CPU
2.Random seed #61
3.CNN accuracy: 0.9757 Perceptron Accuaracy: 0.9190

1. T4 GPU
2. Random Seed: 92
3. Single Layer Perception: 0.8395, Multi-Layer Perception: 0.8808, CNN: 0.9218
   Google Colab

1. T4
2. Random seed = 66
3. CNN accuracy = 0.9305; MLP accuracy = 0.8736

With Random seed 77, CNN had accuracy of 0.9945
Also tried the A100 GPU, which trained the CNN in 83s vs. 123s for T4.

1. T4 GPU
2. 57
3. Single Layer: 0.8413
   Multi-Layer: 0.8781
   CNN: 0.9281
   [Google Colab]

1. T4 GPU
2. 36

• Perceptron Accuracy: 0.8483
• MLP accuracy: 0.8832
• CNN accuracy: 0.9251

1.T4 GPU
2. Random seed: 18
3. CNN: 92.99%, Perceptron: 84.38%, MLP: 88.28%

1. T4 GPU
2. random seed 19
3. SLP: .8410
   CNN:.9310
   MLP: .8838

1. T4 GPU
2. Random seed: 74
3. 1. CNN accuracy: 0.9247
   2. Perceptron accuracy: 0.8437

1. T4 GPU
2. Seed: 84
3. Perception: 0.8452
   MLP: 0.8842
4. Colab: Google Colab

1. v6e-1 TPU
2. Random Seed: 36
3. CNN accuracy: 0.9034
   Perceptron Accuracy: 0.8352
   MLP accuracy: 0.8505

Copy of Lab3NNForFashionMNIST.ipynb - Colab

1. T-4 High Ram.
2. My random seed was 96.
3. Perceptron: 84%
   MLP: 87.33%
   CNN: 92.6%

1. T-4 GPU
2. Random seed was 45.
3. Perceptron: 84.04%
   MLP: 88.71%
   CNN: 92.64%