Different problems to be solved with deep learning

There are many problems which we can solve with deep learning.

We will categorize the problems.

One to One

Each sample in the dataset consists of a single input and it’s corresponding class or label.

We take one image of a cat and label it as “cat”. Similarly do that for the whole dataset.

Image Recognition

image → “cat”

One to Many

Each sample in the dataset consists of a single input and the output consists of a sequence.

Image Captioning

Many to One

Each sample in the dataset consists of sequence input and the single output.

Sentiment Analysis

“it was a good movie” → “positive”

Many to Many(Encoder — Decoder)

Each sample in the dataset consists of sequence input and the sequence output. Number of time steps in the input and output need not be equal.

Language Translation, Chatbots

Many to Many

Each sample in the dataset consists of sequence input and the sequence output. Number of time steps in the input and output should be equal.

Next word prediction, Part of Speech Tagging.