15846588
Description
Flashcards by August Edström, updated more than 1 year ago
|
Created by August Edström
over 6 years ago
|
|
| Question | Answer |
| What is machine learning? | Methodology for onstructing models to predict properties of data |
| What is the difference between supervised and unsupervised learning | The knowledge of the output values(known/unknown) |
| What is special about reinforcement learning? | The model learns behavior in an environment |
| How is the output of a node determined? | By the activation funtion |
| How is the error handled in a neural network? | Backpropagation |
| How can input be encoded? | Dummy coding or vector representation(word embedding) |
| Why is input encoding needed? | to be able to represent a large number of category values in a relatively compact input space |
| What does the distributional hypothesis say? | It states that words that often occur in a similar context are semantically similar |
| A standard neural network has a single set of weights, a recurrent neural network (RNN) has two sets of weights, and a Long Short Term Memory (LSTM) network has three sets of weights. What is the purpose of the extra sets of weights for the RNN and LSTM models? | for the RNN, the extra weights are used to incorporate the output of the previous element in a sequence into the output of the current element. For the LSTM, the third set of weights is used to maintain a cell state, which models long term dependencies beyond just the previous element. |
| What is the purpose of a word embedding in text classification? | word embeddings are used as a dimension reduction technique, to be able to represent a large number of category values in a relatively compact input space. |
| Given the words 'pizza', 'pasta' and 'recursion', construct 5-dimensional word embeddings with two significant digits for each word such that the semantic similarity between the words is preserved. | for example pizza = [0.46,0.88,0.21,0.23,0.52], pasta = [0.43,0.81,0.77,0.19,0.52], recursion = [0.12,0.03,0.54,0.98,0.80]. |
Want to create your own Flashcards for free with GoConqr? Learn more.