-
Notifications
You must be signed in to change notification settings - Fork 11
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Time and time series core concept improvements #84
Merged
+70
−22
Merged
Changes from all commits
Commits
Show all changes
2 commits
Select commit
Hold shift + click to select a range
File filter
Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
|
@@ -57,25 +57,3 @@ If a time is not specified, the resulting recommendation query time will default | |
[Training Rewards]({{<ref "reference/pod#rewards">}}) are code definitions in Python that tell the Spice.ai AI Engine how to train the neural networks to achieve the desired goal. A reward is defined for each action specified in the pod. | ||
|
||
In the future we will expand the languages we support for writing the reward functions in. [Let us know](mailto:[email protected]) which language you want to be able to write your reward functions in! | ||
|
||
## Time | ||
|
||
Time is a fundamental building block for the Spice.ai project. Spice.ai can natively understand how to process time series data, whether that data is streamed in a continuous manner or is batched into larger timeframes. | ||
|
||
Taking a pod manifest and creating multiple pods with different time parameters will result in pods that can learn how to give recommendations for actions at different time frames. (i.e. should an action be taken once every second or once every day) | ||
|
||
### Epoch | ||
|
||
An epoch defines the beginning, or start, of the data stream. If Spice.ai receives data from before the epoch time, it is not used during training. If the epoch is omitted, the epoch is inferred to be `now` - `period`. | ||
|
||
### Period | ||
|
||
A period is the total span of time that is considered for a pod. The end of the data stream that Spice.ai will look at is the `epoch` + `period`. | ||
|
||
### Interval | ||
|
||
The interval is the time span that Spice.ai uses as a single input to the neural networks that power Spice.ai. Attempting to get a recommendation without Spice.ai having at least one intervals worth of data will result in an error. | ||
|
||
### Granularity | ||
|
||
The granularity is the smallest unit of time that specifies how many timesteps there are in an interval. The granularity cannot be larger than the interval. When streaming data in a continuous manner, the Spice.ai runtime can give a new recommendation for action after each new granularity's worth of data is collected. |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,70 @@ | ||
--- | ||
type: docs | ||
title: "Time" | ||
linkTitle: "Time" | ||
weight: 5 | ||
description: "The concept of Time and Time-Series in Spice.ai" | ||
--- | ||
|
||
Spice.ai is a time series AI platform, so time is one of Spice.ai's most fundamental and core concepts. | ||
|
||
## What is time series data? | ||
|
||
Time series data is a series of timestamped data points or events indexed in time order. In Spice.ai, these data points are called Observations. For example, temperature sensor readings on an interval or a daily stock price are examples of time series data. Time series data applies to many domains, including analytics, finance, health and biometrics, IoT and industrial, security, and application monitoring. To learn more, Wikipedia has a comprehensive article on [time series](https://en.wikipedia.org/wiki/Time_series), and InfluxData has published a very informative article, [What is time series data?](https://www.influxdata.com/what-is-time-series-data/) | ||
|
||
## Why time series AI? | ||
|
||
One definition of artificial intelligence proposed by [Hado van Hasselt from DeepMind](https://www.youtube.com/watch?v=TCCjZe0y4Qc&list=PLqYmG7hTraZDVH599EItlEWsUOsJbAodm) is "to be able to learn to make decisions to achieve goals." Using this definition can conclude that these decisions are made over time to achieve an application's goals. Thus, time and time-series data can be seen as fundamental concepts when developing an application's intelligence. | ||
|
||
## Time series in Spice.ai | ||
|
||
Spice.ai natively ingests, processes, and learns from time series data. There are three core time series concepts in Spice.ai: | ||
|
||
- A time series **Period** and its period **Epoch** | ||
- Consecutive **intervals** or windows of time series observations | ||
- The smallest **granularity** of time in the series | ||
A visualization of these concepts over a timeline is below: | ||
|
||
<img width="782" alt="spiceai-time" src="https://user-images.githubusercontent.com/80174/144559519-fc875d71-b43f-402a-ba2e-6754ebba9df0.png"> | ||
|
||
Each of these concepts translates to developer configurable parameters in the Spicepod. For example, the Trader Sample uses these pod parameters in its manifest: | ||
|
||
```yaml | ||
name: trader | ||
params: | ||
period: 30m | ||
interval: 30s | ||
granularity: 5s | ||
``` | ||
|
||
If not provided in the manifest, Spicepods will default to a period of **3 days**, intervals of **1 min**, and granularity of **10 seconds**. The period epoch will default to a dynamic epoch of the current time minus the period. In this mode, the period becomes a sliding window over time. | ||
|
||
See reference documentation for [Spicepod params]({{<ref "reference/pod#params">}}). | ||
|
||
### Period | ||
|
||
The `period` defines the entire timespan the Spicepod will use for learning and decision-making. | ||
|
||
Thus the time series period has a start time of `epoch` and an end time of `epoch` + `period`. | ||
|
||
### Period Epoch | ||
|
||
The period `epoch` defines the beginning, or start, of the Spicepod's time series data. Spice.ai will discard data timestamped before the epoch time. | ||
|
||
The epoch defaults to the current time - `period`. | ||
|
||
### Interval | ||
|
||
The `interval` defines the interval or window of time the AI engine uses to learn. | ||
|
||
The first interval of data, from `epoch` to `epoch` + `interval` is considered pre-training or "warm-up" data. At least one interval's worth of data is required before the AI engine can learn and make a decision recommendation. | ||
|
||
The AI engine then trains on consecutive data intervals for the rest of the period to learn. | ||
|
||
### Granularity | ||
|
||
The `granularity` defines the smallest unit of time series data. The granularity can be conceptualized in practice as the time a decision is valid. For example, a decision to turn on or off an air conditioner would be valid for a `granularity` unit of time. | ||
|
||
Spice.ai will aggregate observations within a granularity timespan. This aggregation results in granularity-sized "ticks" or time-steps in the series. | ||
|
||
Each consecutive training interval advances by a unit of granularity over the entire period. Thus, the granularity cannot be larger than the interval. |
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.