> For the complete documentation index, see [llms.txt](https://huang-jason.gitbook.io/deep/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://huang-jason.gitbook.io/deep/symbolic-reasoning.md).

# Symbolic Reasoning

<https://deeplearning4j.org/symbolicreasoning>

## Two flaws of deep learning

1. lack of model interpretability (i.e. why did my model make that prediction?) \[solution: deep symbolic learning]
2. the amount of data requires in order to learn. They are data hungry. \[solution: one-shot learning]

## Symbol and Sign

The sign or symbol is a visual pattern, string of characters, in which meaning is embedded.

EX: Rose, which is pointing at the red, curling petals layered one over the other in a tight spiral at the end of a stalk of thorns.

**Symbols compress sensory data in a way that enables humans, those beings of limited bandwidth, to share information.**

## Reasoning

Combinations of symbols that express their interrelations could be called reasoning

## symbolic reasoning = expert systems

Implementations of symbolic reasoning are called rules engines or expert systems or knowledge graphs

## Difference = where the learning happens

One of the main differences between machine learning and traditional symbolic reasoning is where the learning happens.

DL: Learns rules as it establishes correlations between inputs and outputs.

SR: the rules are created through human intervention.

## Problems with Symbolic AI

1. Additional rules can’t undo old knowledge. \[but DL can be retrained on new data.]
2. Computers do not know what the symbols mean. \[DL links symbols to vectorized representations of the data]

## DL + SR

In supervised learning, those strings of characters are called labels, the categories by which we classify input data using a statistical model.