Motivations: Neural Networkd [Non-linear hypotheses]

Non-linear Hypotheses

今天是傳說中的 7/8 日!! 哈哈哈

• Motivation- Non-linear Classification:
  • 如果今天 features 有一堆!
  • ex: 房子/房價

      x1 = size,        # quadratic features (o(n^2)) about: (n^2)/2  [n --> features]
      x2 = # bedrooms   ====> ≈ 5000 features
      x3 = # floors   
      x4 = age          
      .                 # cubic features (o(n^3))
      .                 ====> 170,000 features
      .                 
      x100
    

  • for many machine learning problems: n will be pretty large

• Example n will be pretty large:
  • Compouter Vision: Car detection
  • EX:

     50 x 50 pixel images -> 2500 pixels
     n = 2500 (7500 if RGB)
           
              |   pixel 1 intensity   | ---> (0 ~ 255)
              |   pixel 2 intensity   |
         x =  |         .             |
              |         .             |
              |  pixel 2500 intensity |
                    
     # Quadratic features (xi * xj): ≈ 3 million features
    

當 features 超級多的時候，用 logistic regression 不是一棒棒的方法去學 complext nonlinear hypothese.

Neurons and the Brain

Neural Networks

• Origins: Algorithms that try to mimic the brain.
• Was very widely used in 80s and early 90s; popularity diminised in late 90s.
• Recent resurgence: State-of-the-art technique for many applications

不論送入甚麼訊號給大腦，大腦就是有能力可以去學習處裡它!!!