CNS Seminar: Week

JavaScript isn't enabled in your browser, so this file can't be opened. Enable and reload.

CNS Seminar: Week

Transformers as a model of human conceptual system

Student ID *

What are the components of a self-attention layer?

1 point

Value

Key

Lock

Query

One way Transformers based networks segregate inputs in an image is?

1 point

By finding and segregating each object instance in the data (represent each object in image, using RCNN, Monet etc))

By segregating each type of object in the data ( i.e. grouped by features)

Find only one key object in the image

Input the representation of image as a whole (i.e. from average pool or FC layer of a CNN)

Clear selection

According to the presentation what is still lacking in Transformers as a model of concepts is that,:

1 point

They cannot take cues from context.

They can take cues from context but not from very long range (like in LSTMs).

They don't have well defined top-down feedback mechanism.

The still don't have the ability to adapt to a new task when trained on other (use the already learnt concepts).

Clear selection

Transformers suffer from the superposition-catestrophy problem ( i.e. are they not able to represent two objects in one image)?

1 point

True

False

Clear selection

Submit

Clear form

Never submit passwords through Google Forms.

This content is neither created nor endorsed by Google. Report Abuse - Terms of Service - Privacy Policy