For problem sets with just nominal feature sets, vector space classification algorithms often fail. What are your thoughts on the statement? How would you use ML to categorise nominal data? Answer the questions above in detail.

If the issue set consists entirely of nominal characteristics, vector space-based classification algorithms are usually not the best option. Let me know what you think about this assertion. How would you apply ML to the task of categorizing nominal data?

You want to build a text classifier that takes in a sequence of text and outputs its sentiment. You want to use a pretrained model to extract a fixed-size representation from the input text sequence, and train a linear classifier on this representation. You would like this representation to capture information from the entire input sequence. A 10 token sequence (including start/stop tokens) is provided as input to the models below. Mark true for the representations that could satisfy this property. The 5th embedding in the final layer of embeddings of BERT. The 10th embedding in the second layer of embeddings of BERT. The 5th embedding in the final layer of embeddings of a unidirectional Transformer. The 10th embedding in the second layer of embeddings of a unidirectional Transformer.

What is true about a validation set?   its data should be similar to that of the training set   we train our models on the validation set to show that they are valid   the validation set is often smaller than the training set   validation sets, while popular, are not necessary to produce ML models to be used in the world   the validation set is the collection of y-values to the inputs to your model, which are the x-values