Artificial Intelligence and Machine Learning
AI and ML
Technical debt is a metaphor which refers to the implied cost of additional work which happens due to having implemented a quick and simplistic machine learning (ML) solution. Some typical design choices which contribute to technical debt are the project changing requirements, outdated code, insufficient unit and system testing and poor ML documentation. Code refactoring ... Read more
Tensor flow TensorFlow is an open-source machine learning (ML) software library. Tensorflow was originally created by the Google Brain Team in 2015. The main purpose of Tensorflow is building and training neural networks, by using a variety of supported languages, the primary one being Python. To discover the Tensorflow customizable pre-built models, visit its official ... Read more
The threshold is a value which a classification model uses, in order to classify anything higher than the threshold as positive, and anything lower than the threshold as negative.
In machine learning, time series is a series of data, where values of certain features are presented in a sequence of time. There are univariate and multivariate time series in forecasting problems in machine learning. Univariate forecasting models make use of algorithms such as ARIMA and multivariate forecasting models make use of algorithms such as ... Read more
TLU stands for threshold logic unit. TLU is an output neuron which calculates the weighted sum of input neurons and then implements a step function. This is used in the perceptron artificial neural network model.
Tokenization in Natural Language Processing (NLP) is the process of partitioning natural language text into smaller units, which are then manipulated by RNN or other artificial neural network models.
Tensor Processing Units (TPUs) are electronic circuits which are optimized for AI operations and ML tasks. TPUs utilize the TensorFlow protocol logic to align with AI/ML task performance requirements.
Transformer machine learning model A transformer is a deep learning model. Transformer models are mainly used for natural language processing (NLP) and computer vision (CV). Transformers are the evolution of RNN models. A recent example of tranformer-type models in artificial intelligence (AI) are the dp-tranformer models developed by Microsoft research. More details can be found ... Read more
Underfitting is a problem in machine learning in which a model cannot make effective target feature value estimations due to an inability to identify the underlying patterns in the data. An underfitting ML model has low variance and high bias. Underfitting is one pole away from a good fit, the other pole being overfitting.
In machine learning, univariate generally means that a property of a dataset has only a single variable which is being studied. Univariate is used in forecasting models, where only a single variable is being considered in a single time series. A common univariate ML algorithm is ARIMA.
VAR stands for vector autoregression. It is a regression algorithm commonly used in multivariate time series forecasting problems. See the ARIMA algorithm for univariate time series forecasting.
In machine learning (ML), variance is a concept which is related to errors in the model's predictions, as a results of over-sensitivity and high correlation of the machine learning algorithm to the training data. Due to this over-sensitivity, the ML model becomes complex to explain (explainability) and it captures the complexity inside the training data ... Read more
WatsonX is IBM's new generation AI and ML platform. It is built on top of the IBM Watson platform and comprises various modules and subsystems, which can be summarized at high-level as follows: Watsonx AI, including AI/ML model preparation, training, evaluation and tuning tools (https://www.ibm.com/products/watsonx-ai) Watson Data, a data lake-based data platform for hosting AI ... Read more
In audio processing, a waveform is a visual representation of signal amplitude over time.
WCSS stands for within-cluster sum of squares. WCSS is a clustering model evaluation metric that measures the compactness of clusters.
In machine learning word embedding refers to a method in natural language processing (NLP) which converts words into numerical vectors. The numerical vectors generally represent the words' meaning, usage or context. The following word embedding methods are the most common: One-hot encoding assigns a unique binary vector to each word in a vocabulary. The vector ... Read more
Z-score in data science is also known as standardization score. Z-score is the number of standard deviations that a sample is above or below the mean of all values in the sample.