Thread by @rohanpaul_ai

1/ Can you classify something without seeing it before - that's what Zero-Shot Learning is all about - A Thread

👉 One of the popular methods for zero-shot learning is Natural Language Inference (NLI).

#DataScience #DeepLearning #MachineLearning #100DaysOfMLCode #Pytho

2/ Zero-shot learning is when we test a model on a task for which it hasn't been trained.

#DataScience #MachineLearning #100DaysOfMLCode #DataScientist #Statistics #programming #ArtificialIntelligence #Data #pythoncode #AI #Stats #DeepLearning #100DaysOfCode #Database #BigData

3/ In Zero-shot classification, we ask the model to classify a sentence to one of the classes (label) that the model hasn't seen during training.

#DataScience #MachineLearning #100DaysOfMLCode #DataScientist #Statistics #programming #ArtificialIntelligence #Data #pythoncode #AI

4/ Natural language inference is the task of determining whether a "hypothesis" is true (entailment), false (contradiction), or undetermined (neutral) given a "premise"

#DataScience #MachineLearning #100DaysOfMLCode #DataScientist #Statistics #programming #ArtificialIntelligence

5/
Using the NLI method, we can propose a sentence to be classified as a Premise and can construct a hypothesis for each classification label
👇

#DataScience #MachineLearning #100DaysOfMLCode #DataScientist #Statistics #programming #ArtificialIntelligence #Data #pythoncode #AI

6/ E.g., Let's say we have the sentence "I love my job about as a Travel Agent" and we would like to classify whether this sentence is about

1. travel,
2. Profession,
3. Being in Love

#DataScience #MachineLearning #100DaysOfMLCode #DataScientist #Statistics #programming

7/
Now for all three classification labels, we can have three hypothesis

This text is about..

Hypothesis 1: Travel.
Hypothesis 2: One's Profession.
Hypothesis 3: Being in Love.

#DataScience #MachineLearning #100DaysOfMLCode #DataScientist #Statistics #programming

8/ 👉 Using a pre-trained NLI model, we can get the probability for each hypothesis given the premise "one day I will see the world."

#DataScience #MachineLearning #100DaysOfMLCode #DataScientist #Statistics #programming #ArtificialIntelligence #Data #pythoncode #AI #Stats

9/
There are two common approaches used to solve the zero-shot recognition problems.

1. Embedding based approach

The goal here is to map the image features and semantic attributes into a common embedding space using a projection function, which is learned using deep networks

10/ 2. Generative model-based approach - Here we generate image features for non-observed categories using semantic attributes.

#DataScience #MachineLearning #100DaysOfMLCode #DataScientist #Statistics #programming #ArtificialIntelligence #Data #pythoncode #AI #Stats

11/ Generally, this is done using a conditional generative adversarial network (cGAN) that generates image features conditioned on the semantic attribute of a given category.

#DataScience #MachineLearning #100DaysOfMLCode #DataScientist #Statistics #programming

12/ Here is a simple implementation of zero shot classification with Hugging Face Pretrained model of Transformer

#DataScience #MachineLearning #100DaysOfMLCode #DataScientist #Statistics #programming #ArtificialIntelligence

For regular tips and techniques on #DeepLearning, #ComputerVision and #MachineLearning follow me @rohanpaul_ai