Help
Options
Didn't know it?
click below
click below
Knew it?
click below
click below
Don't Know
Remaining cards (0)
Know
retry
shuffle
restart
0:00
Embed Code - If you would like this activity on your web page, copy the script below and paste it into your web page.
Normal Size Small Size show me how
Normal Size Small Size show me how
Ling Final
| Question | Answer |
|---|---|
| How do humans compensate for invariance? | Knowing the rules of language/society Ignoring 'unimportant' variation Working around hard to understand phonemes |
| What is categorical perception? | Ignoring 'unimportant' variation |
| What is phoneme restoration? | Using word context to determine which phonemes were heard and the phonemes to determine which word was heard |
| How do humans solve variation? | Rules of Dialectology Rules of Phonological Rules Categorical Perception Context |
| What is dialectology? | Using the rules of the language/culture to decide what to ignore |
| What is rule based speech recognition? | Human expert inputs: Phonological rules, Dialectology rules, Coarticulation rules, etc |
| What is neural network AI? | Model finds patterns for input-output mappings without a human |
| Deep learning formula? | W.x + b = a |
| What is adjusted in stochastic gradient descent? | Weights, biases, activation function |
| What is the process of diffusion? | Input 1: object with progressively more noise Input 2: text associated with object Output: object ex: speech or image |
| What is diffusion? | Model is trained to find the object through the noise by eliminating irregularity from the signal, eliminating noise by making each point in the signal more similar to its neighbor |
| How does generative speech AI work? | Trained on 2 inputs Noisy speech Text Output: clean speech |
| What are the types of neural network artificial intelligence systems? | Diffusion Large Language Model |
| How does a large language model work? | Each NN’s output provides the input to the next NN |
| How does machine learning work? | Data Driven Automatic Doesn’t require a programmer to tell it the rules Does require a programmer to provide it with training input-output sets |
| What kind of data do you need for machine learning? | More Data = More Accurate Representative of the users Initial training sets: Labeled for what the desired outputs are |
| What are tools for deep learning? | Stochastic gradient descent The chain rule Backpropagation |
| How does stochastic gradient descent work? | Learn by adjusting knowledge of mappings (weights) a little at a time (with y<1) over multiple comparisons with real data points chosen at random new estimate = old estimate - y(Error) |
| How does the chain rule work? | Describes how the weight for one layer affects the overall activation of the network |
| What is backpropagation? | The most efficient route through the network when learning accurate weights |
Created by:
user-1886742