Knowledgebase

Viewing articles tagged 'machine'

Algorithms and Modelling: Everything That Matters, Briefly The summary.FIT A LINEAR OR LOGISTIC MODEL FIRST AND RECORD ITS PERFORMANCEIt is fast,... Applied Use Cases: Everything That Matters, Briefly The summary.THE DEFINITION OF THE TARGET IS THE PROJECTChurn, default, failure and relevance each... Assessing Whether You Have Usable Data Data readiness.WHAT TO ESTABLISHWhether examples of the outcome exist, historically How many... Bias, Variance and Error Analysis Understanding where error comes from.WHAT BIAS ISError from a model too simple to represent the... Building a Machine Learning Portfolio Demonstrating capability.WHAT DEMONSTRATES LITTLETutorial reproductions Well-known teaching... Building an ML Platform Infrastructure for model work.WHAT IT SHOULD PROVIDEAccess to data Environments for... Building Prediction Interfaces Serving requests.WHAT THE INTERFACE MUST DOAccept input Validate it Preprocess it Predict... Building With Large Language Models Engineering around language models.WHAT THE OPTIONS ARE, IN ORDER OF COSTPrompting a hosted model... Classification Metrics Measuring categorical predictions.WHAT THE CONFUSION MATRIX SHOWSCorrect positives, correct... Classifying and Routing Support Requests Machine learning in support operations.WHAT THE TASKS ARECategorising by topic Assessing urgency... Content Moderation Systems Detecting prohibited content.WHAT THE TASKS AREClassifying content against policy categories... Continuous Delivery for Machine Learning Automating the path to production.WHAT DIFFERS FROM SOFTWARE DELIVERYThree things change, not... Convolutional Networks Architectures for images and grids.WHAT A CONVOLUTION DOESApplies a small learned filter across... Credit and Risk Scoring Assessing likelihood of repayment.WHAT THE TARGET ISDefault, defined by a period and a... Customer Churn Prediction Predicting who will leave.WHAT MUST BE DEFINED FIRSTWhat churn means.WHAT THE OPTIONS AREExplicit... Decision Trees and Ensembles The workhorses of tabular data.WHAT A DECISION TREE DOESSplits data repeatedly on feature values,... Deep Learning Frameworks Building neural networks.WHAT THE MAIN OPTIONS AREA framework favouring imperative, Python-native... Deep Learning: Everything That Matters, Briefly The summary.MOST WORK IS ADAPTING AN EXISTING ARCHITECTURE, NOT INVENTING ONEStart from a... Delivering Machine Learning Projects Making them succeed.WHAT KILLS PROJECTSNo clear decision the prediction supports Data that does... Demand and Sales Forecasting Predicting quantities ahead.WHAT TO ESTABLISHThe horizon: how far aheadThe granularity: per... Deploying Computer Vision Models Vision systems in production.WHAT THE PIPELINE INVOLVESImage capturePreprocessing: resizing,... Deploying Recommendation Systems Operating recommenders in production.WHAT THE SERVING PATTERN USUALLY ISCandidate generation,... Deployment Patterns for Models How predictions reach users.WHAT THE PATTERNS AREBatch scoring, written to a table An interface... Designing Honest Evaluation Estimates you can trust.WHAT UNDERMINES AN ESTIMATETuning against the test set Selecting the best... Detecting and Preventing Leakage The error that ruins everything.WHAT LEAKAGE ISInformation in training data that would not be... Development Environments and Notebooks Where the work is done.WHAT NOTEBOOKS PROVIDEInteractive exploration Output alongside code... Distance-Based and Probabilistic Methods Other classical approaches.WHAT NEAREST NEIGHBOUR METHODS DOPredict from the most similar... Document and Form Processing Extracting structure from documents.WHAT THE STAGES AREClassification: what kind of documentText... Documenting Models Responsibly Recording what a model is and is not.WHAT TO DOCUMENTIntended use, and uses considered... Evaluating Language Model Systems Measuring generated output.WHY IT IS DIFFICULTThere is no single correct output, and quality is... Evaluation and Training: Everything That Matters, Briefly The summary.EVERY CLASSIFICATION METRIC DERIVES FROM THE CONFUSION MATRIXStart there. Precision... Experiment Tracking and Model Versioning Keeping track of what was tried.WHAT TO RECORD FOR EVERY EXPERIMENTThe code version The data... Experimenting With Model Changes Testing whether a change helps.WHY OFFLINE IMPROVEMENT IS INSUFFICIENTA better metric does not... Fairness in Machine Learning Systems Models that treat people equitably.WHY IT MATTERSModels make or influence decisions about people,... Feature Selection and Dimensionality Choosing what the model sees.WHY IT MATTERSIrrelevant features add noise, cost and opportunity... Framing a Machine Learning Problem Turning a business need into a task.WHAT TO ESTABLISHWhat decision is being made Who or what... Fraud and Abuse Models Detecting bad activity.WHAT MAKES IT DISTINCTIVEThe adversary adapts Positives are rare Decisions... Handling Different Data Types Preparing inputs.WHAT NUMERICAL FEATURES MAY NEEDScaling, for distance and penalty-based methods... Handling Imbalanced Data When one outcome is rare.WHY IT MATTERSA model predicting the majority class always can appear... Handling Model Failure Gracefully When prediction is unavailable or wrong.WHAT CAN FAILThe model service being unavailable Features... Human Oversight of Automated Decisions Keeping people in the loop.WHAT OVERSIGHT MODELS EXISTA person decides, with the model advising A... Hyperparameter Tuning Finding good settings.WHAT A HYPERPARAMETER ISA setting chosen before training, not learned from... Interpreting Models Explaining predictions.WHY IT MATTERSDecisions affecting people require explanation Debugging... Keeping Current Without Chasing Everything Sustainable learning.WHAT THE PROBLEM ISThe volume of new work vastly exceeds what anyone can... Labelling Data Producing the answers to learn from.WHERE LABELS COME FROMRecorded outcomes, which are best Human... Learning and Career: Everything That Matters, Briefly The summary.LEARN EVALUATION EARLY, BEFORE FRAMEWORKSIt is where beginners most often go wrong,... Learning Machine Learning Properly Building real competence.WHAT THE COMMON MISTAKE ISLearning frameworks before fundamentals.WHY... Linear and Logistic Models The baseline worth beating.WHAT LINEAR REGRESSION DOESFits a weighted sum of features to predict... Machine Learning Engineering: Everything That Matters, Briefly The whole category in one page.GENERALISATION IS THE ONLY THING THAT MATTERSTraining performance... Machine Learning Foundations: Everything That Matters, Briefly The summary.GENERALISATION IS THE ONLY THING THAT MATTERSPerformance on training data is... Machine Learning Roles Compared Who does what.WHAT A DATA SCIENTIST TYPICALLY DOESAnalysis, experimentation and modelling,... Machine Learning Work in the Nigerian Context Local realities.WHAT THE OPPORTUNITIES AREProblems with local data nobody else has addressed... Managing Environments and Dependencies Reproducible setups.WHY IT MATTERS MORE HERE THAN ELSEWHERENumerical libraries have tight version... Managing Model Cost Controlling spend.WHAT DRIVES COSTTraining runs Inference volume Hardware type Data storage and... Mathematics for Machine Learning The foundations that matter.WHAT LINEAR ALGEBRA PROVIDESThe language of the field: vectors,... Model Deployment: Everything That Matters, Briefly The summary.PREFER BATCH SCORING WHERE IT FITSNo latency requirement, no serving infrastructure,... Monitoring and MLOps: Everything That Matters, Briefly The summary.MONITOR INPUT DISTRIBUTIONS, BECAUSE OUTCOMES ARRIVE LATE OR NEVERActual performance... Monitoring Models in Production Knowing whether it still works.WHAT TO MONITOROperational health: availability, latency,... Neural Network Fundamentals How they work.WHAT A NEURON COMPUTESA weighted sum of inputs, plus a bias, passed through a... Optimising Inference Performance Making prediction fast and affordable.WHAT TO MEASURELatency, at the percentiles that matter... Packaging Models for Deployment Making a model runnable.WHAT MUST BE PACKAGEDThe model artefact The preprocessing applied to... Predictive Maintenance Anticipating equipment failure.WHAT IT PREDICTSThat equipment will fail within a window, or its... Preparing for Machine Learning Interviews What is assessed.WHAT IS TYPICALLY COVEREDFundamentals: evaluation, overfitting, algorithm... Prioritising Machine Learning Work Choosing what to build.WHAT TO ASSESS FOR EACH CANDIDATEThe value if it works The probability it... Privacy in Machine Learning Protecting people in training data.WHAT THE RISKS AREModels memorising training examples... Probability Calibration When predicted probabilities must be believable.WHAT CALIBRATION MEANSAmong predictions of... Problem Framing and Data: Everything That Matters, Briefly The summary.DEFINE THE TARGET PRECISELY BEFORE ANYTHING ELSEPredicting churn requires defining... Propensity and Pricing Models Predicting response and willingness.WHAT A PROPENSITY MODEL PREDICTSLikelihood of a particular... Python for Machine Learning The working language.WHY IT DOMINATESThe libraries, overwhelmingly.WHAT THE CORE LIBRARIES... Reading Research Papers Keeping up with the literature.WHY IT MATTERSThe field moves quickly, and much useful work... Recommendation Modelling Predicting preferences.WHAT THE APPROACHES AREContent-based: similarity between... Regression and Ranking Metrics Measuring numerical and ordered predictions.WHAT MEAN ABSOLUTE ERROR MEASURESAverage size of... Releasing Models Safely Controlled deployment.WHAT THE STRATEGIES AREShadow deployment: running alongside, predictions... Responsible and Applied Practice: Everything That Matters, Briefly The summary.REMOVING PROTECTED ATTRIBUTES DOES NOT MAKE A MODEL FAIROther features correlate with... Retraining Strategy Keeping models current.WHAT THE OPTIONS AREOn a schedule When performance degrades When drift is... Scikit-learn in Practice The classical machine learning library.WHAT IT PROVIDESConsistent interfaces across many... Search and Ranking Models Ordering results by relevance.WHAT LEARNING TO RANK DOESLearns an ordering from examples of what... Selecting Between Models Choosing what to deploy.WHAT TO COMPARE ONPerformance on the metric that matters Performance on... Sequence Models and Attention Architectures for ordered data.WHAT RECURRENT NETWORKS DIDProcessed sequences step by step,... Splitting Data for Evaluation Separating training from assessment.WHAT THE SPLITS ARETraining: what the model learns... Supervised Learning Fundamentals Learning from labelled examples.WHAT IT REQUIRESExamples, each with input features and a known... Testing Machine Learning Systems Verification before deployment.WHAT TO TEST ABOUT THE CODEPreprocessing, against known inputs and... The Machine Learning Lifecycle How projects actually proceed.WHAT THE STAGES AREFraming the problem Establishing whether data... Time Series Modelling Predicting over time.WHAT MAKES IT DIFFERENTObservations are ordered and dependent The future... Tools and Environment: Everything That Matters, Briefly The summary.ALWAYS USE A PIPELINE OBJECT FOR PREPROCESSINGIt is what prevents leakage, since... Training Deep Learning Models Practical training.WHAT THE LOOP ISForward pass, compute loss, compute gradients, update... Training Infrastructure and Cost Running training practically.WHAT DETERMINES COSTHardware type and hours Data transfer Storage of... Transfer Learning and Fine-Tuning Starting from an existing model.WHAT IT ISUsing a model trained on a large general dataset as the... Understanding Drift Why models decay.WHAT DATA DRIFT ISThe input distribution changing.WHAT CONCEPT DRIFT ISThe... Understanding Generative Models Models that produce rather than classify.WHAT THEY DOLearn a distribution and sample from it.WHAT... Understanding Model Capacity and Regularisation Controlling how much a model can learn.WHAT CAPACITY ISHow complex a pattern a model can... Unsupervised and Self-Supervised Learning Learning without labels.WHAT UNSUPERVISED LEARNING DOESFinds structure in data with no known... Using Pre-trained Model Repositories Starting from published models.WHAT THEY PROVIDEPre-trained weights Standardised loading... Versioning Data and Models Keeping track of what produced what.WHAT MUST BE VERSIONEDCode Data Model artefacts Environment... When Not to Use Machine Learning Recognising the wrong tool.WHAT A RULE SOLVES BETTERAnything with clear, stable, enumerable... Working as a Machine Learning Engineer The role in practice.WHAT THE WORK ACTUALLY ISUnderstanding problems and whether they are... Working With Accelerated Hardware Training on graphics processors.WHAT DETERMINES WHAT YOU CAN TRAINMemory, primarily.WHAT CONSUMES... Working With Image and Audio Data Practical handling.WHAT IMAGE PREPROCESSING INVOLVESResizing to the model's expected dimensions... Working With Text Data Natural language processing practice.WHAT TOKENISATION DOESSplits text into units the model...
Back

Are you happy with your experience? Leave us a review on Trustpilot.


Trustpilot