Artificial Intelligence and Machine Learning Flashcards
What is a transformer models role in generative ai
Able to process a sentence as a whole instead of word by word.
Faster and more efficient text processing
Relative importance given to specific words in a sentence
Generate human like text
What is deep learning ?
Uses neurons and synapses to train a model.
Process more complex patterns in the data that traditional ML
Deep Learning is made up of multiple input, hidden and output layers
Examples are computer vision, NLP.
Requires larger amounts of data and GPU processors
What are neural networks ?
Nodes are organised in layers and as data is pushed into the model then new connections are made between these nodes and sometimes some node connections are removed. Nodes talk to each other forming new connections. Neural networks may have billions of nodes.
What are hyper parameters ?
A hyperparameter are the settings that define the model structure and learning algorithm and process. These are set before training begins.
Give some examples of hyper parameters
Learning Rate
Batch Size
Number of Epochs - Number of iterations
Regularisation -
What are the two types of inferencing ?
Real Time and Batch
What is batch inferencing ?
Large amounts of data that is analysed all at once. Often used for data analysis. Speed of the results is usually not a concern but accuracy is
What is real time inferencing ?
This is where computers have to make decisions quickly as data arrives. Speed is preferred over accuracy. An example is that of chatbots.
What are the two models for inferencing on edge
SLM on edge device or LLM accessed via api
If I wanted low latency, low impact inferencing at the edge what model would I choose ?
SLM
What metrics should I use on a binary classification.
Confusion Matrix
What are the four metrics for a confusion matrix
Precision - Best when false positives are costly
Recall - Best when false negatives are costly
F1 - Best when you want a balance between precision and recall especially in imbalanced datasets
Accuracy - Best for balanced data sets
Name the regression metrics ?
Mean Absolute Error
Mean Absolute Percentage Error
Root Mean Squared
R2
What is the difference between R2 and the other regression metrics
R2 is for variance the others are accuracy metrics
What is the Area under the curve metric
AUC-ROC shows what the curve for true positive compared to false positive looks like at various thresholds with multiple confusion matrixes. A value nearer to 1 is better
What is GPT
generate human text or code based on input prompts
What is Bert ?
similar to GPT but reads text in two directions
What is RNN ?
meant for sequential data such as timer series or text useful in speech recognition or time series data
What is Wave Net ?
Model to generate raw audio waveform used in speech synthesis
What are the four layers of AI
Data Layer - collection of vast amounts of data
ML Framework and Algorithm Layer - data scientists and engineers work together to understand use cases, requirements and frameworks
Model Layer - Implementation of a model and train it.
Application Layer - How to serve the model and its capabilities for your users.
