Artificial Intelligence and Machine Learning
AI and ML
Accuracy in machine learning classification model (classifier) is a measure of how frequently each classification is correctly deemed positive or negative. Accuracy is calculated by the following mathematical formula accuracy = (true positives + true negatives) / (all estimated values) Use the following reference for some good visual examples of accuracy, precision and recall: https://www.evidentlyai.com/classification-metrics/accuracy-precision-recall.
An activation function in artificial neural networks (ANN) is a function which computes the output of an artificial neuron to solve non-linear tasks. The activation function of a neural network essentially decides whether a neuron should be activated or not.
The term adversarial machine learning describes all possible defense measures against cyber attacks on machine learning models.
The Adversarial ML Threat Matrix is a framework aimed at detecting and resolving cybersecurity threats in ML systems. The Adversarial ML Threat Matrix is based on the MITRE ATT&CK Matrix.
AI accelerator An AI accelerator is a specialized hardware component which is mainly utilized for artificial intelligence tasks. It can perform AI computations in a more efficient manner than conventional computer processors (CPU), hence AI accelerators are also known as Neural Processing Units (NPU). By using algorithms for parallelism, AI accelerators can enchance performance of ... Read more
Artificial Intelligence (AI) relates to computer hardware in multiple ways, as follows: AI platforms are being used to optimize the design and manufacturing of new compute hardware systems. Special computer hardware processor units, referred to as AI accelerators or Neural Processing Units (NPU) are being developed to enhance the performance of AI applications and machine ... Read more
An algorithm is a series of instructions, either in the form of pseudo-code or in a computer programming language, which aim at solving a problem or at performing some sort of computation. Algorithms are therefore strongly connected to various fields of mathematics and physics, such as probability and statics, calculus and linear algebra. Algorithms apply ... Read more
ANN stands for artificial neural network. It is an electronic neural network which simulates the operations and features of human brain neural networks.
ARIMA stands for autoregressive integrated moving average. It is a regression algorithm which is commonly used in univariate time series forecasting problems.
Artificial intelligence (AI) means developing intelligence in computing systems and robots. Artificial intelligence emulates human brain intelligence and is capable of executing tasks which require cognitive ability. The most common cognitive services which can be implemented in AI are the following: Language, including Natural Language Processing (NLP) and Natural Language Understanding (NLU) Computer vision, including ... Read more
Application-Specific Integrated Circuits (ASICs) are custom-built integrated circuit (IC) electronic systems, which are designed to optimize AI and ML workloads. ASICs have a large application spectrum, including for example video and voice codecs. In contrast to ASICs which are application specific, there are various general purpose Integrated Circuit systems, such as FPGAs.
AUC stands for Area Under the Curve. This refers to the two-dimensional area which is defined by a mathematical plot curve, the x-axis and two boundary points.
In data engineering, augmentation is a process by which we create various transformations of the available data in the ML training dataset. Augmentation task examples are the perturbation of an image in different ways. Augmentation aims at increasing the amount of data input provided to a model to better fit the data into the model ... Read more
In machine learning (ML), bias is a concept which is related to errors in the model's predictions, as a results of multiple assumptions and simplifications in the machine learning algorithm. Due to these assumptions, the ML model becomes easy to explain (explainability) but it often misses to capture the complexity inside the training and testing ... Read more
Bias in machine learning is a type of error which occurs when an ML model's estimations are different than the ground truth, i.e what is known (and validated) to be correct.
Caffe is a deep learning framework, which has been developed by Berkeley AI Research (BAIR) and various other contributors. The Caffe official project website is https://caffe.berkeleyvision.org/.
CNN stands for convolutional neural network. It is a type of artificial neural network which is primarily used in artificial intelligence for image processing and computer vision tasks. CNN network architecture comprises an input layer, followed by one or more pairs of convolution and pooling layers connected in series, then followed by fully connected MLP ... Read more
The coefficient of determination (also depicted as R^2) is a calculation of the proportion of the variance in the dependent variable (noted as y) which is explained by a linear regression model. The value of the coefficient of determination can be in the 0..1 range. The coefficient of determination can be calculated by using the ... Read more
A convex function is a function which features a single global minimum, whereas a non-convex functions presents many local minima. A good analysis on the mathematical aspects and mathematical definition of convex and non-convex functions can be found at: https://rumn.medium.com/convex-vs-non-convex-functions-why-it-matters-in-optimization-for-machine-learning-39cd9427dfcc. Some examples of ML algorithms which have restrictions related to convex functions are the following: ... Read more
The loss function (or sometimes called error function) is a function which compares the output of an ML model as compared to the ground truth for a single training example, while the cost function is meant to be calculated over the entire training set (or mini-batch for mini-batch gradient descent).