Scribd Logo
Upload

Welcome to Scribd!

  • 94 views

    Machine Learning Models: by Mayuri Bhandari

    Uploaded by

    mayuri
    The document discusses different types of machine learning models including supervised learning, unsupervised learning, reinforcement learning, and semi-supervised learning. It then describes concepts related to evaluating machine learning models such as bias, variance, underfitting, overfitting, and the bias-variance tradeoff. Finally, it covers evaluation metrics for classification problems including accuracy, confusion matrices, F1 score, and for regression problems including mean absolute error, mean squared error, and maximum likelihood estimation.

    Copyright:

    © All Rights Reserved
    Download as PPTX, PDF, TXT or read online from Scribd

    Machine Learning Models: by Mayuri Bhandari

    Uploaded by

    mayuri
    94 views48 pages
    The document discusses different types of machine learning models including supervised learning, unsupervised learning, reinforcement learning, and semi-supervised learning. It then describes concepts related to evaluating machine learning models such as bias, variance, underfitting, overfitting, and the bias-variance tradeoff. Finally, it covers evaluation metrics for classification problems including accuracy, confusion matrices, F1 score, and for regression problems including mean absolute error, mean squared error, and maximum likelihood estimation.

    Original Title

    2.Machine Learning Models

    Copyright

    © © All Rights Reserved

    Available Formats

    PPTX, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    The document discusses different types of machine learning models including supervised learning, unsupervised learning, reinforcement learning, and semi-supervised learning. It then describes concepts related to evaluating machine learning models such as bias, variance, underfitting, overfitting, and the bias-variance tradeoff. Finally, it covers evaluation metrics for classification problems including accuracy, confusion matrices, F1 score, and for regression problems including mean absolute error, mean squared error, and maximum likelihood estimation.

    Copyright:

    © All Rights Reserved
    Download as PPTX, PDF, TXT or read online from Scribd
    Download as pptx, pdf, or txt
    94 views48 pages

    Machine Learning Models: by Mayuri Bhandari

    Uploaded by

    mayuri
    The document discusses different types of machine learning models including supervised learning, unsupervised learning, reinforcement learning, and semi-supervised learning. It then describes concepts related to evaluating machine learning models such as bias, variance, underfitting, overfitting, and the bias-variance tradeoff. Finally, it covers evaluation metrics for classification problems including accuracy, confusion matrices, F1 score, and for regression problems including mean absolute error, mean squared error, and maximum likelihood estimation.

    Copyright:

    © All Rights Reserved
    Download as PPTX, PDF, TXT or read online from Scribd
    Download as pptx, pdf, or txt
    of 48
    At a glance
    Powered by AI
    The document discusses different types of machine learning including supervised, unsupervised, reinforcement and semi-supervised learning.

    The types of machine learning discussed are supervised learning, unsupervised learning, reinforcement learning and semi-supervised learning.

    Supervised learning involves providing a machine with labeled training data to perform a specific task. The machine predicts answers and receives feedback on its predictions to learn.

    Machine Learning Models

    By Mayuri Bhandari
    Types of Machine Learning
    Supervised Learning
    Supervised learning is when you provide the machine
    with a lot of training data to perform a specific task.
    In the training data, you’d feed the machine with a lot
    of similar examples, and the computer will predict the
    answer.
    You would then give feedback to the computer as to
    whether it made the right prediction or not. 
    Supervised learning is task-specific, and that’s why it’s
    quite common. 
    Supervised Learning
    Unsupervised Learning
    As the name suggests, unsupervised learning is the
    opposite of supervised learning.
    In this case, you don’t provide the machine with any
    training data. 
    The machine has to reach conclusions without any
    labeled data.
    It’s a little challenging to implement than supervised
    learning.
    It is used for clustering data and for finding anomalies.
    It is also quite popular as it is data-driven.
    Unsupervised Learning
    Reinforcement Learning
    Reinforcement learning is quite different from other types of
    machine learning (supervised and unsupervised). 
    The relation between data and machine is quite different
    from other machine learning types as well. 
    In reinforcement learning, the machine learns by its
    mistakes.
    You give the machine a specific environment in which it can
    perform a given set of actions.
    Now, it will learn by trial and error.
    Although reinforcement learning is quite challenging to
    implement, it finds applications in many industries. 
    Reinforcement Learning
    Semi-supervised Learning
    A semi-supervised learning problem starts with a
    series of labeled data points as well as some data point
    for which labels are not known. 
    The goal of a semi-supervised model is to classify
    some of the unlabeled data using the labeled
    information set.
    The goal of a semi-supervised learning model is to
    make effective use of all of the available data, not just
    the labelled data like in supervised learning.
    Semi-supervised Learning
    Machine Learning Models
    Components of Generalization Error

    Bias

    Variance

    Underfitting

    Overfitting
    Bias Error
    Bias is defined as the average squared difference between
    predictions and true values.
    Bias measures the deviation between the expected output of
    our model and the real values, so it indicates the fit of our
    model.
    Bias results in under-fitting the data.
    A high bias means our learning algorithm is missing
    important trends among the features.
    High bias algorithms are easier to learn but less flexible, due
    to this they have lower predictive performance on complex
    problems.
    Bias Error

    Data is almost always noisy in reality, so some bias is


    inevitable — called the irreducible error
    Variance
    Variance measures the amount that the outputs of our
    model will change if a different dataset is used.
    It is the impacts of using different datasets.
    A model is said to have high variance if its predictions
    are sensitive to small changes in the input. 
    Generally, non-parametric machine learning algorithms
    that have a lot of flexibility have a high variance
    Bias-Variance Trade-off
    Example : Bias-Variance Trade-off
    Bias-Variance Trade-off

    Training dataset 1 Training dataset 2


    High Bias
    High Variance
    How to achieve Trade-off
    Dimensionality Reduction
    Regularisation in linear models
    Using mixture model and Ensemble training
    Optimal value of K in KNN
    Total Error
    Error = Bias + Variance
    Total Error = Bias + Variance + Irreducible error
    error(X) = bias(X)2 + variance(X) + noise(X)
    Bias(X) = E[f^(x)] − f(x)
    Underfitting and Overfitting
    Underfitting and Overfitting
    Overfitting: Good performance on the training data,
    poor result while giving other data.

    Underfitting: Poor performance on the training data


    and poor result while giving the other data.

    Underfitting would imply that the model has still


    capacity to learn, so you would simply train for more
    iterations or collect more data.
    Underfitting and Overfitting
    A learning system cycle
    1. Ideation
    The following prerequisites are essential for a successful
    ideation:
    1. Clear requirements regarding business objectives and
    scope
    2. Availability of historical data
    3. Understanding of end-to-end IT infrastructure
    requirements
    2. Development
    Once key metrics that correspond to the business
    objectives are agreed upon and historical data is acquired,
    the data scientist can start developing the initial model.
    Data scientists have a wide array of tools available to solve
    their puzzles:
    1. Transforming data to a more useful format
    2. Analysis of data to guide modeling approach
    3. Writing of the actual machine learning model code
    4. Creating numbers and visuals for initial reports towards
    stakeholders
    3. Production
    When the development phase is over, the developed model
    needs to be put in production to start generating value.
    The complexity of getting a model in production depends
    on the context of the problem, the autonomy of data
    science teams and the overall maturity of the organization.
    The context of the problem consists of a number of factors:
    1. Data flow at prediction time
    2. Sensitivity of the data
    3. Maximum acceptable latency of delivery
    4. Maintenance
    Once a model is deployed, there are a number of
    measures that can be taken to improve robustness and
    quality of the machine learning model.
    These measures can be roughly divided into four
    areas. We call this post-production process
    maintenance.
    1. Lineage
    2. Monitoring
    3. Comparison
    4. Model Drift
    Evaluation Metrics : Accuracy
    It is the ratio of number of correct predictions to the
    total number of input samples.

    It works well only if there are equal number of samples


    belonging to each class.
     Classification Accuracy is great, but gives us the false
    sense of achieving high accuracy.
    Confusion Matrix
    Confusion Matrix as the name suggests gives us a
    matrix as output and describes the complete
    performance of the model.
    Confusion Matrix
    Lets assume we have a binary classification problem.
    We have some samples belonging to two classes : YES
    or NO. Also, we have our own classifier which predicts
    a class for a given input sample. On testing our model
    on 165 samples ,we get the following result.
    Confusion Matrix
    There are 4 important terms :
    True Positives : The cases in which we predicted YES
    and the actual output was also YES.
    True Negatives : The cases in which we predicted NO
    and the actual output was NO.
    False Positives : The cases in which we predicted YES
    and the actual output was NO.
    False Negatives : The cases in which we predicted
    NO and the actual output was YES.
    Confusion Matrix
    Accuracy for the matrix can be calculated by taking
    average of the values lying across the “main
    diagonal” i.e
    F1 Score
    F1 Score is used to measure a test’s accuracy
    F1 Score is the Harmonic Mean between precision and
    recall. The range for F1 Score is [0, 1].
    It tells you how precise your classifier is (how many
    instances it classifies correctly), as well as how robust
    it is (it does not miss a significant number of
    instances).
    F1 Score : Precision and Recall
    Precision : It is the number of correct positive results
    divided by the number of positive results predicted by
    the classifier.
    True Positives
    Precision =
    True Positives + False Positives
    Recall : It is the number of correct positive results
    divided by the number of all relevant samples 
    True Positives
    Recall =
    True Positives + False Negatives
    Mean Absolute Error
    Mean Absolute Error is the average of the difference
    between the Original Values and the Predicted Values.
    It gives us the measure of how far the predictions were
    from the actual output.
    However, they don’t gives us any idea of the direction
    of the error i.e. whether we are under predicting the
    data or over predicting the data. 
    Mean Squared Error
    Mean Squared Error(MSE) is quite similar to Mean
    Absolute Error, the only difference being that MSE takes
    the average of the square of the difference between the
    original values and the predicted values.
    The advantage of MSE being that it is easier to compute the
    gradient
    As, we take square of the error, the effect of larger errors
    become more pronounced than smaller error, hence the
    model can now focus more on the larger errors.
    Maximum likelihood Estimation(MLE)
    Maximum likelihood estimation is a method that
    determines values for the parameters of a model.
    The parameter values are found such that they
    maximize the likelihood that the process described by
    the model produced the data that were actually
    observed.
    What are parameters?
    For a linear model we can write this as y = mx + c. In this
    example x could represent the advertising spend and y might be
    the revenue generated. m and c are parameters for this model.
    Different values for these parameters will give different lines
    MLE : Example
    Let’s suppose we have observed 10 data points from some process. For
    example, each data point could represent the length of time in seconds
    that it takes a student to answer a specific exam question. These 10 data
    points are shown in the figure below
    MLE : Example
    For these data we’ll assume that the data generation process can
    be adequately described by a Gaussian (normal) distribution.

    Gaussian distribution
    has 2 parameters. The
    mean, μ, and the
    standard deviation, σ.
    Different values of these
    parameters result in
    different curves
    MLE
    We want to know which curve was most likely
    responsible for creating the data points that we
    observed? 
    Maximum likelihood estimation is a method that will
    find the values of μ and σ that result in the curve that
    best fits the data.
    The true distribution from which the data were
    generated was f1 ~ N(10, 2.25), which is the blue curve
    in the figure above.
    Posterior probability
    A posterior probability, in Bayesian statistics, is the
    revised or updated probability of an event occurring
    after taking into consideration new information.
    The posterior probability is calculated by updating
    the prior probability using Bayes' theorem.
    In statistical terms, the posterior probability is the
    probability of event A occurring given that event B has
    occurred.
    Bayes' Theorem Formula
    The formula to calculate a posterior probability of A
    occurring given that B occurred:
    Bayes' theorem can be used in many applications, such as medicine,
    finance, and economics.
    In finance, Bayes' theorem can be used to update a previous belief
    once new information is obtained.
    Prior probability represents what is originally believed before new
    evidence is introduced, and posterior probability takes this new
    information into account.
    Posterior probability distributions should be a better reflection of the
    underlying truth of a data generating process than the prior
    probability since the posterior included more information.
     A posterior probability can subsequently become a prior for a new
    updated posterior probability as new information arises and is
    incorporated into the analysis.
    References
    https://google.com
    https://towardsdatascience.com
    https://medium.com
    https://www.upgrad.com/
    www.edureka.co

    You might also like