Zaya Barrini Psychoanalysis, Art and Cinema Studio

Machine Learning

Professional experience

Algorithm Development and Dynamic pricing

Engineering for user recommendations

ML techniques to improve dynamic pricing + maximize profits

Predictive Modeling and Data Mining

Reduce waste, improve sales, find new markets

Text identification - latent words

Supervised model development, testing, and validation

Predict stock market price with high accuracy.

Smart investments

Optimal pricing strategy to achieve revenue goals

Devised high security: detect an abnormality, intrusion, fraud, masquerading, malware.

Customer behavior.

Rate the financial competence of the business

Personalized application with large data

Predict product sales

Problem-Solving Process

Identify data that is relevant to the problem

Assemble a set of data related to the problem you’re trying to solve

Decide on the type of output you are predicting

Based on the type of output, pick an algorithm that will determine a correlation between your features and labels

Use a model generated by an algorithm to make a prediction

Classification - Labels belong to a discreet set

Regression - labels belong to a continuous set

Features == independent variable

Label == dependent variable

Untitled content

Algorithm

KNN - K-nearest Neighbor - which container will the ball fall into?

Training data

Test data

Learning

Predict results

Accuracy

Learn from tests

Shuffle data, Random

Statistics

Gauging accuracy

Investigating optimal values

Our prediction was bad

Adjust the parameters of the analysis

Add more features to explain it

Change the prediction point

Accept that maybe this isn’t a good correlation

KNN for multiple inputs

Performance Issues

Feature Normalization, Standardization

Feature Selection

Evaluating different features - accuracy

Linear Regression

Only train one time - use it for any prediction

Multiple independent variables

Gradient Descent

Feature Normalization, Standardization

Learning rate

Mean squared error

Derivative

y = mx + b

m is called slope

Slope isthe derivative

Algorithm: pick a value for b and m.

Calculate the slope of MSE with respect to b and m

Are both slopes very small? If so, we are done

Multiply both slopes by learning rate

Subtract results from b and m

Learning rate, iterations, features, labels, options

Train the model

Use the test to make predictions about observations with known labels

Gauge accuracy

Multivariate First degree equation

Learn rate optimization methods: Adam, Adagrad, RMSProp, Momentum

Batch Gradient Descent - subset of data

Stochastic Gradient Descent - one row at a time

Natural Binary Classification

Logistic Regression

The sigmoid equation

Cross-Entropy

Cost functions

Multinomial Classification

Marginal vs Conditional Probability

Sigmoid vs Softmax

argMax

Handwriting Recognition

Multinomial Logistic Problem

Image file - pixels

..flapMap - reduce 1 dimension in space. 2D to 1D

Encoding Labels values

debugger

Optimization

Handling Large Datasets

Memory issue

Memory Allocation

Memory snapshot

Handling Large Datasets

Minimizing Memory Usage

Creating Memory Snapshots

The Javascript Garbage Collector

Shallow vs Retained Memory Usage

Measuring Memory Usage

Releasing References

Measuring Footprint Reduction

Optimization Tensorflow Memory Usage

Tensorflow’s Eager Memory Usage

Cleaning up Tensors with Tidy

Implementing TF Tidy

Tidying the Training Loop

Measuring Reduced Memory Usage

One More Optimization

Final Memory Report

Plotting Cost History

NaNin Cost History

Fixing Cost History

Massaging Learning Parameters

Improving Model Accuracy

Multinominal Classification

Marginal vs Conditional Probability

Sigmoid vs Softmax

argMax

Algorithms in Lacanian Psychoanalysis

Representation: - Graphs, Venn Diagrams, Schemes - Pseudocode - Programming code…

Representation: - Graphs, Venn Diagrams, Schemes - Pseudocode - Programming code

Scheme of the signifying chain and the production of meaning - Something is left out (Phenomenology …

Scheme of the signifying chain and the production of meaning - Something is left out (Phenomenology of the spirit + The critique of pure reason) - Defect of enjoyment - Scheme of the Tori - Entangling of Tori === Unconscious

I, unconscious, Subject of the unconscious - Speech, who speaks, to whom it speaks - Response, multi…

I, unconscious, Subject of the unconscious - Speech, who speaks, to whom it speaks - Response, multi-purpose interpretation - Chance and Interpretation - Chance and Love - Chance and finding the lost object - Reencounter - Prediction, Inference - Repetition, Loops - Neurotic Calculus of Enjoyment

Seminars [1,3]: Imaginary, especially ethology and optics, Mirror Stadium…

Seminars [1,3]: Imaginary, especially ethology and optics, Mirror Stadium

Seminars [4,8]: Symbolic, Oedipus complex, French structuralism, Claude Lévi-Strauss’s anthropology and Saussure’s linguistics

Seminars [9,20]: Real, Theory of alienation, logic with topology, mathematics: the limits of language

Seminars [21,27]: RSI, knot theory

Untitled content

![](/images/Psychoanalysis-Courses-Concepts-Machine-Learning-Psychoanalysis.pptx/Psychoanalysis-Courses-Concepts-Machine -Learning-Psychoanalysis_5.png)

Untitled content

![](/images/Psychoanalysis-Courses-Concepts-Machine-Learning-Psychoanalysis.pptx/Psychoanalysis -Courses-Concepts-Machine-Learning-Psychoanalysis_8.png)

Untitled content

![](/images/Psychoanalysis-Courses-Concepts-Machine-Learning-Psychoan alysis.pptx/Psychoanalysis-Courses-Concepts-Machine-Learning-Psychoanalysis_11.png)

Untitled content