agileartificialintelligence / agileartificialintelligence.github.io Goto Github PK

Check the assertion in test of the Class NeuronLayer and NNetwork

By Yvan Guemkam

At the end of section exercise, you state Create a new test, called testSmallExample, that tests that feeding our perceptron p with different values (e.g., -2 and 2 gives 0 as result).

It is not clear what you mean for the user to do.

Should the method be given a values parameter, in which case it should be named testSmallExample:?

Or should it instead be the body of the original testSmallExample method that should be augmented with additional sub-tests, e.g.,

testSmallExample
	| p result |
	p := Neuron new.
	p weights: #(1 2).
	p bias: -2.

	result := p feed: #(5 2).
	self assert: result equals: 1.

	result := p feed: #(-2 2)
	self assert: result equals: 0.

?

A few more observations on Chapter 2 requiring deeper investigation:

• When you introduce the #assert:equals:, it would be great to show a GT
  debugger with the diff pane at the bottom, which gives a hint at what makes
  Pharo special
• Capitalized temps in #digitalComparator:?! Sacrilege ;-)
• "Grapher" is introduced as if the reader already knows what that is. Is
  that assumed to be the case?

By Sean

But i have another problem, when i launch tests on NeuronLayer and NNetwork some assertions fail.
The value #(0.97 ......) not be reached with only 10000 epoch

By Yvan Guemkam