Machine learning

Dagger (Data Aggregation)

A tool used in data processing to combine multiple data sources into a unified dataset for analysis. The data process will initially be off course, and over time adjust to the desired response to task/generation/detection. will have “multiple rounds”. 1 → bad … 10 → good.

Supervised Learning

Trains the model to apply correct outputs to unseen data in real world cases. Using a dataset to understand relations between inputs and outputs. will have a ground truth dataset w/ corresponding labels. Did many projects on this in college.. uses methods such as gradient decent determining numbers vs letters predictablility.

Imitation Learning

Powerful way to implicitly program robots. Robot feedback drives covariate shift between train/test distribution. Uses Data Aggregation

Machine Learning Stage

Predictive models xyzxyz and so on will predict next 3 to be xyz

Deep learning stage

Neurel Network or the BRAIN that will create network of how close in meanings different things mean

Foundation models (LLM/Genai) stage

Uses weighted neurel network to predict the next work, sentance, paragraph, answer using closest idealogical location of weights in the words used and interpretted. placed into words for the ease of communication. Genai is the ability to use building blocks like music notes and words that it understands to fit there with meaning it expects to create the song to create a new song on those building blocks.

ROS communication protocol

Robot Communication Protocol. When it’s a headache to continuously add connections between sensors/actuators/controlls and hardware you can use ROS. Powerful and scalable opperations (Autonomous system w/ gps). Code separation into blocks