DAT301m

What is the key advantage of Long Short-Term Memory (LSTM) networks over traditional Recurrent Neural Network (RNNs) ?

• Faster training times

• Simplicity in implementation

• Better handling of long-range dependencies

• Reduced risk of overfitting

What does Mean Squared Error (MSE) metric measure in the context of forecasting models?

• The average percentage difference between predicted and actual values

• The average absolute difference between predicted and actual values

• The average squared difference between predicted and actual values

• The correlation coefficient between predicted and actual value

What is autocorrelation in the context of time series?

• The correlation between different features in the data

• The correlation between time series

• The correlation between time series and its lagged values

• The correlation between spatially distant observations

What type of tokenization strategy does the Whitespace Tokenizer follow?

• Character-level tokenization

• Subword-level tokenization

• Word-level tokenization

• Setence-level tokenization

Which of following is a common application of time series analysis in finance?

• Image recognition

• Fraud detection

• Stock price forecasting

• Customer segmentation

How does the concept of word embeddings contribute to word-based encodings?

• By reducing words to their root forms

• By converting words into one-hot encoded vectors

• By capturing semantic relationships between words in dense vectors

• Word embeddings have no role in word-based encodings

What is common format for datasets related to dogs and cats on Kaggle?

• CSV files

• XML files

• JPEG images

• MP3 audio files

How does the SentencepieceTokenizer handle a rare of out-of-vocabulary words?

• It replaces them with a special token

• It ignores them during tokenization

• It treats each character as a separate token

• It creates subword units for them based on learnt patterns.

How does a pre-trained model contribute to achieving good results with a small training dataset?

• It completely replaces the need for any training on the small dataset

• It provides an initial set of weights that capture generic features, enhancing model performance

• It hinders the model's ability to adapt to the small dataset

• Pre-trained models are note suitable for small dataset

How can you add a GPU to your Colab notebook for faster TensorFlow computations?

• Purchase a Colab Pro subscription

• It is not possible to use GPUs in Colab

• Click on "Runtime" and then select "Change runtime type"

• Use a third-party GPU provider and integrate it with Colab