Featuring pointers for: Single-layer neural networks and multi-layer neural networks, gradient descent, backpropagation. Slides are for introduction, for deep explanation on deep learning, please consult other slides.
2. Organic Neural Network
• The human brain has about 1011 neurons
• Switching time 0.001s (computer ≈ 10-10s)
• Connections per neuron: 104 - 105
• 0.1s for face recognition!
• Strengths: Parallelism and distributedness
3. Biological Neurons
• Dendrit menerima input informasi dalam bentuk sinyal elektrik,
yang diakumulasikan di badan sel saraf (= soma).
• Ketika akumulasi informasi mencapai threshold tertentu,
sel saraf akan menembakkan output informasi
yang dikirim melalui axon
• Axon tersambung dengan dendrit yang ada
di badan sel saraf lain melalui synapses
• Pemelajaran/learning dilakukan dengan
adaptasi synaptical weight
5. ANN: Basic Idea
• Artificial Neuron
• Each input is multiplied by a weighting factor
• Output is: 1 if sum of weighted inputs exceeds threshold;
0 otherwise
• Network is programmed by adjusting weights using feedback from
examples
17. ANN Learning Using Gradient Descent
For an excellent step-by-step tutorial on Gradient Descent:
https://mccormickml.com/2014/03/04/gradient-descent-derivation/
22. Backpropagation
1. Computes the error term for the output units using the observed error.
2. From the output layer, repeat:
• propagating the error term back to the previous layer, and
• updating the weights between the two layers
until the earliest hidden layer is reached.
43. Neural network in action: AlphaGo
More info about CNN:
https://towardsdatascience.com/the-most-intuitive-and-easiest-guide-for-convolutional-neural-network-3607be47480
44. Neural network in action: Self-driving car
More info about CNN:
https://towardsdatascience.com/the-most-intuitive-and-easiest-guide-for-convolutional-neural-network-3607be47480
Diasumsikan untuk node i
Penjelasan:
Input diterima 'neuron'
Input diakumulasi dengan input function
Melalui fungsi aktivasi, dihasilkan output a_i
Fungsi aktivasi $g$ bisa berupa fungsi sigmoid, fungsi step/threshold, dsb
Neural network merupakan kumpulan unit atau node (dari unit input hingga unit output) yang terhubung dan membentuk topologi neuron
Step function = binary
Sigmoid function = non-binary (70% activated, 10% activated, etc)
Rectified Linear Unit
https://analyticsindiamag.com/most-common-activation-functions-in-neural-networks-and-rationale-behind-it/
Input bersifat high-dimensional.
Memodelkan relasi yang non-linear dan kompleks.
Proses untuk mendapatkan hasil (atau interpretability) tidak penting = black box
a0 = -1
We assume step function = 0 if x < 0 and 1 otherwise