chrishayuk/simple-classifier

Data Generation

the following will generate data for the simple binary classifier

python data_generator.py --num_samples 2000  --input_size 10

Train the model

python train.py

Evaluate the model

python evaluate.py

activation function

this uses a silu activation function. started with ReLU but switched to SiLU for better results

dropout layer

this has a dropout layer set to 0.1 probability, applied after each hidden layer. this will essentially drop a neuron during training 10% of the time, to prevent the model overfitting. it helps the model not rely on particular neuron configurations dropout is not applied during inference