Initialization on the How to Train Models track. Initial weights must break symmetry and keep early activations in a sensible range. Framework defaults are usually right. Random huge weights waste the start of training.
This lesson assumes you already worked through Learning-rate schedules.
The idea in practice
Use the default initializer for the layer type. Change it only when you can show a failure it fixes.
A concrete check
goal = {
'track': 'How to Train Models',
'lesson': 'Initialization',
}
checks = [
'input available at decision time',
'score matches the real decision',
'failure case written down',
]
print(goal['lesson'])
for item in checks:
print('-', item)
Run the sketch locally if you have Python. The printout is a reminder of the checks, not a trained model. Replace the strings with the real inputs from your own example before you treat it as a design.
What usually goes wrong
Zeroing every weight in a deep net so every neuron computes the same thing. When this happens, stop adding parameters or tools. Fix the check, the data, or the permission, then run the same example again.
What to write down
- The input you are allowed to use at decision time.
- The output and the score or pass rule.
- One failure you will test on purpose.
- What you will not claim the system can do.
Practice
Explain why all-zero weights fail in a layer with more than one neuron.
Self-check
- Say Initialization in one sentence that mentions an input and an output.
- Name the failure mode in this lesson and the check that would catch it.
Done when: you can explain this lesson without the page open, and you have a written failure case.