Nick Mitchell mightyhorst

## @learn-d2l.ai-transformers
I see where the confusion might have come from, especially with the introduction of a query value "2" without a clear explanation. Let's clarify the concept of a query in the context of the attention mechanism and adjust the example for better understanding.

In the context of attention mechanisms, a **query** (\(\mathbf{q}\)) represents the current item or piece of information the model is focusing on or trying to relate to other pieces of information. The **keys** (\(\mathbf{k}_i\)) represent aspects or features of other items in a dataset, and the **values** (\(\mathbf{v}_i\)) represent the content or information of those items. The goal of the attention mechanism is to determine how relevant each item (key-value pair) in the dataset is to the query and to produce a weighted combination of these items' values based on their relevance.

Let's correct and simplify the example without introducing arbitrary numbers for the query:

Imagine we have a dataset of items, each with a key (\(\mathbf{k}_i\)) and a val

## DeepMind x UCL RL Lecture Series: 1. Introduction to Reinforcement Learning
Environment - dynamics of the problem
Reward - specifies the goal
Agent -
 * agent state
 * policy
 * value function estimate - optional
 * model - optional

## 1. Inside the agent: state


## MIT 6.S191 - Lecture 2: Recurrent Neural Networks
t is single timestep

y hat of t = function of input vector x at time t

y hat is predicted

temporal dependence

recurence relationship
h is internal state of memeory passed timestep to timestep

## editor.html
<!DOCTYPE html>
<html>
  <head>
    <style>
      .editor { font-family: 'Roboto Mono', monospace; font-size: 12px; outline: none; overflow-y: auto; padding-left: 48px; counter-reset: line; }
      .editor div { display: block; position: relative; white-space: pre-wrap; }
      .editor div::before { content: counter(line); counter-increment: line; position: absolute; right: calc(100% + 16px); opacity: 0.5; }
    </style>
  </head>

## scrollinto.jsx
//from https://css-tricks.com/snippets/jquery/smooth-scrolling/

// Scroll to specific values
// scrollTo is the same
window.scroll({
  top: 2500,
  left: 0,
  behavior: 'smooth'
});

## index.css
//----------------------------------*\
// TRIGONOMETRY FUNCTIONS
//----------------------------------*/

// # Trigonometry in CSS
//
// - Through Taylor/Maclaurin polynomial representation: http://people.math.sc.edu/girardi/m142/handouts/10sTaylorPolySeries.pdf
// - Useful if you don't want to use JS.
// - With CSS Variables.
// - `calc()` can't do power (x ^ y) so I used multiplication instead.

## hello-kubernetes-first.yaml
apiVersion: v1
kind: Service
metadata:
  name: hello-kubernetes-first
spec:
  type: ClusterIP
  ports:
  - port: 80
    targetPort: 8080
  selector:

## jose.service.ts
import {JWK, JWS} from 'node-jose';
import {join} from 'path';
import {writeFileSync, readFileSync} from 'fs';
import fetch from 'node-fetch';

/**
 * @requires Models
 */
import {IdTokenPayload, AccessTokenPayload} from '../models';

## index.ts
/**
 * @namespace @monorepo/common
 * @desc barrel file to export the common files
 */
export * from './models';

## nestjs.package.json
{
  "name": "@monorepo/nestjs",
  "version": "1.0.0",
  "private": true,
  "dependencies": {
    "@monorepo/common": "1.0.0"
  }
}
	I see where the confusion might have come from, especially with the introduction of a query value "2" without a clear explanation. Let's clarify the concept of a query in the context of the attention mechanism and adjust the example for better understanding.

	In the context of attention mechanisms, a query (\(\mathbf{q}\)) represents the current item or piece of information the model is focusing on or trying to relate to other pieces of information. The keys (\(\mathbf{k}_i\)) represent aspects or features of other items in a dataset, and the values (\(\mathbf{v}_i\)) represent the content or information of those items. The goal of the attention mechanism is to determine how relevant each item (key-value pair) in the dataset is to the query and to produce a weighted combination of these items' values based on their relevance.

	Let's correct and simplify the example without introducing arbitrary numbers for the query:

	Imagine we have a dataset of items, each with a key (\(\mathbf{k}_i\)) and a val
	Environment - dynamics of the problem
	Reward - specifies the goal
	Agent -
	* agent state
	* policy
	* value function estimate - optional
	* model - optional

	## 1. Inside the agent: state
	t is single timestep

	y hat of t = function of input vector x at time t

	y hat is predicted

	temporal dependence

	recurence relationship
	h is internal state of memeory passed timestep to timestep
	<!DOCTYPE html>
	<html>
	<head>
	<style>
	.editor { font-family: 'Roboto Mono', monospace; font-size: 12px; outline: none; overflow-y: auto; padding-left: 48px; counter-reset: line; }
	.editor div { display: block; position: relative; white-space: pre-wrap; }
	.editor div::before { content: counter(line); counter-increment: line; position: absolute; right: calc(100% + 16px); opacity: 0.5; }
	</style>
	</head>
	//from https://css-tricks.com/snippets/jquery/smooth-scrolling/

	// Scroll to specific values
	// scrollTo is the same
	window.scroll({
	top: 2500,
	left: 0,
	behavior: 'smooth'
	});
	//----------------------------------*\
	// TRIGONOMETRY FUNCTIONS
	//----------------------------------*/

	// # Trigonometry in CSS
	//
	// - Through Taylor/Maclaurin polynomial representation: http://people.math.sc.edu/girardi/m142/handouts/10sTaylorPolySeries.pdf
	// - Useful if you don't want to use JS.
	// - With CSS Variables.
	// - `calc()` can't do power (x ^ y) so I used multiplication instead.
	apiVersion: v1
	kind: Service
	metadata:
	name: hello-kubernetes-first
	spec:
	type: ClusterIP
	ports:
	- port: 80
	targetPort: 8080
	selector:
	import {JWK, JWS} from 'node-jose';
	import {join} from 'path';
	import {writeFileSync, readFileSync} from 'fs';
	import fetch from 'node-fetch';

	/**
	* @requires Models
	*/
	import {IdTokenPayload, AccessTokenPayload} from '../models';
	/**
	* @namespace @monorepo/common
	* @desc barrel file to export the common files
	*/
	export * from './models';
	{
	"name": "@monorepo/nestjs",
	"version": "1.0.0",
	"private": true,
	"dependencies": {
	"@monorepo/common": "1.0.0"
	}
	}