albers

Data ScienceComplete with Branch IllustrationPhoto by Praveen Thirumurugan on UnsplashYou’re a data scientist. As data science is becoming more and more mature every day, software engineering practices begin creeping in. You are forced to venture out of your local jupyter notebooks and meet other data scientists in the wild to build a great product.To help you out with this grand mission, you can rely on Git, a free and open-source distributed version control system to keep track of what everyone is coding.To be more…

Deep LearningExplaining their complex mathematical formula with working diagramsPhoto by Rubaitul Azad on UnsplashTable of Contents· Recurrent Neural Networks (RNN)∘ Vanilla RNN∘ Long Short-term Memory (LSTM)∘ Gated Recurrent Unit (GRU)· RNN Architectures· Attention∘ Seq2seq with Attention∘ Self-attention∘ Multi-head Attention· Transformer∘ Step 1. Adding Positional Encoding to Word Embeddings∘ Step 2. Encoder: Multi-head Attention and Feed Forward∘ Step 3. Decoder: (Masked) Multi-head Attention and Feed Forward∘ Step 4.…

Data ScienceWhy Word2Vec is better, and why it’s not good enoughPhoto by Tamara Bellis on Unsplash, edited by authorWhile image data is straightforward to be used by deep learning models (RGB value as the input), this is not the case for text data. Deep learning models only work on numbers, not sequences of symbols like texts. So, you need a way to somehow extract meaningful numerical feature vectors from texts. This is called feature extraction.From now on, we will call a single observation of text by document and a…

Deep LearningUnderstanding CNN from the ground upImage by authorTable of Contents· Fully Connected Layer and Activation Function· Convolution and Pooling Layer· Normalization Layer∘ Local Response Normalization∘ Batch Normalization· 5 Most Well-Known CNN Architectures Visualized∘ LeNet-5∘ AlexNet∘ VGG-16∘ Inception-v1∘ ResNet-50· Wrapping UpThe introduction of LeNet in 1990 by Yann LeCun sparks the possibility of deep neural networks in practice. However, limited computation capability and memory capacity made the…

Random Forest is one of the most popular bagging methods. What’s bagging you ask? It’s an abbreviation for bootstrapping + aggregating. The goal of bagging is to reduce the variance of a single estimator, i.e. the variance of a single Decision Tree in the case of Random Forest.To be concrete, let’s use a dummy dataset throughout this story. Suppose you have the following tax evasion dataset. Your task is to predict whether a person will comply to pay taxes (the Evade column) based on features like Taxable Income (in…

Machine LearningWith hands-on case study walkthroughsPhoto by Julian Hochgesang on UnsplashThere are 2 major kinds of machine learning models: supervised and unsupervised. In supervised learning, you have input data X and output data y, then the model finds a map from X to y. In unsupervised learning, you only have input data X. The goal of unsupervised learning varies: clustering observations in X, reducing the dimensionality of X, anomaly detection in X, etc.As supervised learning has been discussed extensively in Part…

Machine LearningWith hands-on case study walkthroughsPhoto by Fabrice Villard on UnsplashUnlike linear models and SVM (see Part 1), some machine learning models are really complex to learn from their mathematical formulation. Fortunately, they can be understood by following a step-by-step process they execute on a small dummy dataset. This way, you can uncover machine learning models under the hood without the “math bottleneck”.You will learn three more models in this story after Part 1: K-Nearest Neighbors (KNN), Naive…

How does linear regression relate to logistic regression and Support Vector Machine?Table of contents· So, how does machine learning work?· Linear Models∘ Linear regression∘ Ridge∘ Lasso∘ Elastic-Net· Logistic Regression· Support Vector Machine (SVM)∘ Classification∘ Regression∘ Kernel functions· A note on preprocessing· ConclusionMachine learning modeling is a data scientist’s problem solver. Even though it doesn’t take the majority of our time, it’s personally much more fun than data cleaning.To be frank, some models…

Hands-on TutorialsLabel your data programmaticallyPhoto by Swanson Chan on UnsplashTable of Contents:· Exploratory Data Analysis· Keyword Labeling Functions· Heuristic Labeling Functions· Labeling Functions with spaCy· Combining Labeling Function Outputs· Training a Classifier· Wrapping UpThere was a radical idea to entirely eliminate hand-labeling any training data in machine learning projects. It birthed snorkel, a powerful library to programmatically build training data.There are three programmatic operations in…

The CRISPR-Cas9 gene-editing technique can be used to change genetic coding with great precision, and its uses include curing disease and fixing genetic faults. Now it's been used to change animal behavior – and the results are not what scientists were expecting.  CRISPR was used on a group of hamsters to knock out the Avpr1a receptor, which is acted on by the hormone vasopressin. Vasopressin is linked to social actions, such as bonding, cooperation, communication, dominance, and aggression.Without the Avpr1a receptor,…