numpy-NN

Neural Network implementation from scratch using numpy

Usage

numpy-nn

model = Network()
model.add(DenseLayer(6))
model.add(DenseLayer(8))
model.add(DenseLayer(3))

model.train(X_train=X, y_train=y, epochs=200)

EPOCH: 0, ACCURACY: 0.3333333333333333, LOSS: 1.8507288360616592

EPOCH: 20, ACCURACY: 0.64, LOSS: 0.8984484293696664

EPOCH: 40, ACCURACY: 0.5666666666666667, LOSS: 0.8055846210908157

EPOCH: 60, ACCURACY: 0.5933333333333334, LOSS: 0.7544998303196496

EPOCH: 80, ACCURACY: 0.6466666666666666, LOSS: 0.7034754660535022

EPOCH: 100, ACCURACY: 0.8666666666666667, LOSS: 0.6522870909240465

EPOCH: 120, ACCURACY: 0.9466666666666667, LOSS: 0.6051327850621049

EPOCH: 140, ACCURACY: 0.96, LOSS: 0.5624822108029988

EPOCH: 160, ACCURACY: 0.96, LOSS: 0.5237726663927962

EPOCH: 180, ACCURACY: 0.9533333333333334, LOSS: 0.4887972804949555

keras equivalent

from keras.models import Sequential
from keras.layers import Dense
import tensorflow as tf
from tensorflow.keras.optimizers import SGD

ohy = tf.keras.utils.to_categorical(y, num_classes=3)

model2 = Sequential()
model2.add(Dense(6, activation='relu'))
model2.add(Dense(10, activation='relu'))
model2.add(Dense(8, activation='relu'))
model2.add(Dense(3, activation='softmax'))

model2.compile(SGD(learning_rate=0.01), loss='categorical_crossentropy', metrics=['accuracy'])

model2.fit(x=X, y=ohy, epochs=30)

Metrics

Notes

• Currently:
  • relu activation in hidden layers
  • softmax activation in the output layer
  • cross-entropy loss
  • multi-class tasks ONLY

Next Steps

1. Documentation
2. Binary classification tasks
3. Regression tasks
4. Multiple loss functions / optimizations