Explanations 📖

A list of concepts and mental models, explained in very simple terms, that help improve critical thinking and decision-making:

Pareto Principle 📈

Approximately 80% of effects stem from 20% of causes, implying that a small number of inputs yield a majority of the results.

Regret Minimization Framework 😖

Imagine yourself in the future, reflecting on your current decision. Choose the option that minimizes potential regret.

This concept ties in nicely with Problem Inversion: the practice of thinking through problems in reverse.

Forced serendipity ☘️

Consistent hard work and dedication increase the likelihood of encountering fortunate events. "The harder I work, the luckier I seem to get."

Principal-Agent Problem 👑

A disagreement in priorities arises between the asset owner and the appointed manager, leading to potential conflicts of interest.

📖 Skin in the Game by Nassim Nicholas Taleb.

Second-Order Thinking 🙋‍♂️

Consider not only the immediate consequences of a decision but also its subsequent effects.

In other words, always ask “And then what?”

Opportunity Cost ⚖️

The potential benefit or value forfeited when choosing one alternative over another.

Barbell strategy 🏋️‍♂️

Balancing risk and reward by investing in high-risk and no-risk assets, while avoiding moderate-risk options.

It's a numbers game 💯

Success often depends on measurable key performance indicators (KPIs) that quantify results.

To Care and Communicate 📢

Achieving success often requires genuine passion for your work and effective communication of your ideas.

Network Effects 🕸

The value of a good or service increases as more people use or participate in it.

Ockham's Razor ✂️

When faced with multiple explanations, the simplest one is often the most likely to be correct.

Parkinson's Law 🕰️

Work expands to fill the time available for its completion, implying that setting shorter deadlines can improve productivity.

Dunning-Kruger Effect 🤦‍♂️

The cognitive bias where individuals with limited knowledge or competence overestimate their abilities, while experts may underestimate their expertise.

NP vs P ⏱

If the solution to a problem is easy to check for correctness, must the problem be easy to solve? (For example, cracking a password).

The general class of questions for which some algorithm can provide an answer in polynomial time is "P". For some questions, there is no known way to find an answer quickly, but if one is provided with information showing what the answer is, it is possible to verify the answer quickly. The class of questions for which an answer can be verified in polynomial time is NP, which stands for "nondeterministic polynomial time”.

Cognitive Biases 🧠

Systematic errors in thinking that affect human behavior and perception of reality.

Groupthink 🐑: The tendency for individuals within a group to conform to a single decision, often resulting in irrational outcomes.

More biases over here.

Logical Fallacies 🤔

Flawed reasoning or invalid deductive arguments that hinder effective decision-making.

Strawman 💀: Misrepresenting an opponent's argument to make it easier to attack.

Sunk Cost Fallacy 🕳️:The tendency to continue investing in a decision based on the amount of resources already committed, rather than evaluating the current and future value.

More fallacies over here.

 

Data Science Glossary 👨‍🔬

Machine Learning Models

Linear regression: A statistical model that tries to find the relationship between a dependent variable and one or more independent variables by fitting a linear equation to the observed data. It is used for predicting continuous-valued outputs.

Logistic regression: A classification algorithm used to model the probability of a certain class or event existing, such as pass/fail, win/lose, or healthy/sick. It works by fitting a logistic curve to the given data, which can then be used to predict the probability of binary outcomes.

Decision trees: A non-parametric supervised learning method used for classification and regression. The model learns to make decisions by recursively splitting the input data into subsets based on the values of input features, constructing a tree-like structure.

Random forest: An ensemble learning method that constructs multiple decision trees during training and combines their predictions to produce a more accurate and robust output. It helps to overcome overfitting and improve generalization.

Gradient boosting: A machine learning technique that builds an ensemble of weak prediction models, typically decision trees, and iteratively improves them by minimizing the loss function using gradient descent. It is used for regression and classification problems.

Neural networks (CNN): A type of artificial neural network called Convolutional Neural Network (CNN) is designed to process grid-like structured data, such as images. CNNs are composed of multiple layers that learn to recognize patterns and features in the input data through a process called convolution, making them particularly effective for image recognition tasks.

Clustering (K-Means): An unsupervised learning algorithm that partitions a dataset into K distinct clusters based on similarity. The algorithm works by iteratively assigning data points to the nearest cluster center and updating the cluster centers based on the average of the data points within each cluster.

Dimensionality reduction (PCA, SVD): Techniques used to reduce the number of features in high-dimensional datasets, making it easier to analyze and visualize the data. Principal Component Analysis (PCA) and Singular Value Decomposition (SVD) are two popular methods that work by transforming the original data into a lower-dimensional space while preserving as much of the information as possible.

Time Series: A set of techniques and models used for analyzing and forecasting data that is indexed in time order, such as stock prices or weather data. Time series models consider patterns, trends, and seasonal effects in the data to make predictions about future values.

Support Vector Machines (SVM): A supervised learning algorithm used for classification and regression tasks. SVMs work by finding the optimal hyperplane that best separates the data points of different classes or predicts the target value with the smallest error.

Naive Bayes: A family of probabilistic classifiers based on Bayes' theorem, which assumes that the features are conditionally independent given the class. Despite its simplicity, Naive Bayes is often effective for text classification tasks, such as spam filtering and sentiment analysis.

k-Nearest Neighbors (k-NN): A non-parametric, lazy learning algorithm used for classification and regression tasks. It works by finding the k training samples closest to a new input and predicting the class or value based on a majority vote or weighted average of these neighbors.

Deep Learning (RNN, LSTM, Transformer): A subset of neural networks that focuses on deep architectures with many layers. Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM) networks, and Transformers are popular deep learning models for handling sequential and time-series data, natural language processing, and other complex tasks.

Reinforcement Learning (Q-Learning, DDPG): A type of machine learning that focuses on training agents to make decisions in an environment to maximize cumulative rewards. Q-Learning and Deep Deterministic Policy Gradient (DDPG) are popular reinforcement learning algorithms.

Autoencoders: A type of unsupervised neural network used for dimensionality reduction, feature learning, and anomaly detection. Autoencoders learn to compress input data into a lower-dimensional representation and then reconstruct the original input from this representation.

Generative Adversarial Networks (GANs): A class of deep learning models that consist of two neural networks, a generator, and a discriminator, which are trained simultaneously in a zero-sum game. GANs can generate new data samples that resemble the original data distribution, making them useful for tasks such as image synthesis and data augmentation.

Data Fallacies

Cherry picking: The act of selectively presenting data points or findings that support a specific conclusion, while ignoring or downplaying evidence that contradicts it. This practice can lead to incorrect conclusions and misleading interpretations of data.

Survivorship bias: The error of focusing on the data points or elements that have "survived" a selection process, while overlooking those that did not. This can lead to inaccurate conclusions and predictions, as the survivorship effect may skew the results.

False causality: The assumption that a causal relationship exists between two variables simply because they are correlated. This fallacy is also known as "correlation does not imply causation," as correlation does not always indicate a direct causal relationship.

Sampling bias: The presence of systematic errors in the data that result from a non-random or unrepresentative sampling process. This can lead to incorrect inferences about the population, as the sample may not accurately reflect the characteristics of the larger group.

Observer effect: The phenomenon where the act of observing or measuring a system can change its behavior or properties. In data analysis, this can lead to skewed results if the presence of the observer or data collector influences the data being collected.

Gambler's fallacy: The mistaken belief that past events can influence the probability of future independent events. For example, assuming that a coin toss is more likely to land on heads after a series of tails, despite the fact that each toss is independent and has an equal probability of landing on either side.

Regression towards the mean: The phenomenon where extreme values in a dataset tend to be followed by values closer to the mean. This can lead to incorrect conclusions, such as attributing a change in performance solely to a specific intervention when it may be due to natural variability.

McNamara fallacy: The error of relying solely on quantitative data and metrics to make decisions or evaluate performance, while ignoring qualitative factors that may be difficult to measure. This can lead to suboptimal decision-making and an overemphasis on easily quantifiable variables.

Simpson's paradox: A statistical phenomenon where a trend or relationship between two variables reverses or disappears when the data is aggregated or separated into groups. This paradox highlights the importance of considering the underlying structure of the data when analyzing relationships between variables.

Underfitting vs. Overfitting (Bias-Variance tradeoff): A common issue in machine learning where a model is either too simplistic (underfitting) or too complex (overfitting). Underfitting occurs when a model has high bias and does not capture the underlying patterns in the data, leading to poor performance. Overfitting occurs when a model has high variance and captures noise in the data, leading to poor generalization to new data. Balancing the tradeoff between bias and variance is key to building effective models.

Ecological fallacy: The error of making inferences about individuals based on aggregated data from a group. This fallacy arises when relationships observed at the group level do not necessarily hold true for individuals within that group.

Confounding variable: A variable that is correlated with both the independent and dependent variables in a study, causing a distortion in the observed relationship between them. Failing to account for confounding variables can lead to false conclusions about causality.

Selection bias: The presence of systematic errors in a study due to the non-random selection of participants, variables, or data points. This can lead to incorrect inferences and generalizations, as the selected sample may not accurately represent the population of interest.

Confirmation bias: The tendency to search for, interpret, and recall information in a way that confirms one's preexisting beliefs or hypotheses. This can lead to biased data analysis and conclusions, as researchers may unconsciously favor evidence that supports their views and disregard contradictory evidence.

Availability heuristic: The cognitive bias that leads people to overestimate the importance or likelihood of events based on their availability in memory. In data analysis, this may cause certain events or trends to be given more weight than they deserve, simply because they are more easily recalled.

Anchoring effect: The cognitive bias that occurs when an initial piece of information is used as a reference point for subsequent judgments and decisions. In data analysis, this can lead to biased estimates and conclusions, as the initial value can unduly influence subsequent evaluations.

Base rate fallacy: The error of ignoring or underestimating the base rate (prior probability) of an event when evaluating the likelihood of that event occurring. This can lead to misinterpretations of probability and overconfidence in the predictive power of specific indicators or pieces of evidence.