README
¶
Examples
This directory contains a few examples of how to use Cofire.
examples.go contains examples of how to start 4 component types:
- The learner consumes ratings from an input topic and applies SGD to learn the latent features of users and products (ie, factorizes the ratings matrix). It also periodically prints to stdout the RMSE of its predictions.
- The producer simply takes a list of ratings and emits them into the learner's input topic.
- The refeeder reemits already learnt ratings into the learner's input topic after a predefined delay and for a number of iterations.
- The validator takes a set of ratings (eg, a test set) and calculates the RMSE using a view on the learnt model (technically, a view of the learner's group table).
Besides these generic starters, there are two concrete examples of recommendation:
- movielens: the classic Movie Lens example, where users are watchers, products are movies and scores are the number of stars each user gives to each movie.
- pixelreco: a funny example that maps the x axis of an image as the users, the y axis as the products and the gray level of the pixel as the score. Any image can be used as input.
Both examples contain the Kafka-based implementation as well as a simple local runner.
Documentation
¶
Index ¶
- func CreateView(brokers []string, group goka.Group) (*goka.View, func(ctx context.Context) func() error)
- func StartLearner(ctx context.Context, brokers []string, group goka.Group, ...) func() error
- func StartProducer(ctx context.Context, brokers []string, group goka.Group, ...) func() error
- func StartRefeeder(ctx context.Context, brokers []string, group goka.Group, delay time.Duration) func() error
- func StartValidator(ctx context.Context, view *goka.View, ratings []cofire.Rating, ...) func() error
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func CreateView ¶
func CreateView(brokers []string, group goka.Group) (*goka.View, func(ctx context.Context) func() error)
CreateView creates a view of the cofire table and a function to start it if no error occurred.
func StartLearner ¶
func StartLearner(ctx context.Context, brokers []string, group goka.Group, params cofire.Parameters) func() error
StartLearner starts a Cofire processor that factorizes a rating matrix with SGD. A simple validator prints to stdout the RMSE every second.
func StartProducer ¶
func StartProducer(ctx context.Context, brokers []string, group goka.Group, ratings []cofire.Rating) func() error
StartProducer starts a producer that emits a slice of ratings into the input of the learner, one rating every 5 milliseconds.
func StartRefeeder ¶
func StartRefeeder(ctx context.Context, brokers []string, group goka.Group, delay time.Duration) func() error
StartRefeeder starts a Cofire refeeder processor, ie, a process that refeeds the stream into the learner's input after a delay. This can be used to train for multiple iterations.
func StartValidator ¶
func StartValidator(ctx context.Context, view *goka.View, ratings []cofire.Rating, params cofire.Parameters) func() error
StartValidator starts a go routine that loops over all given ratings and calculates the RSME calculating the ratings with the error predicted from the model.
Types ¶
This section is empty.
Source Files
Directories