Nacaskul Poomjai 2019.03.31.MachineLearningFundamentals - Reformatted

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/335690621
Machine Learning Fundamentals
Presentation · March 2019
CITATIONS READS
0 5,262
1 author:
Poomjai Nacaskul
Chulalongkorn University
29 PUBLICATIONS 30 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Thailand Sustainable Development Foundation View project
Data Analytics & Machine Learning View project
All content following this page was uploaded by Poomjai Nacaskul on 02 January 2021.
The user has requested enhancement of the downloaded file.

Machine Learning Fundamentals –
Definition & Paradigms, Algorithms &
Languages, Application & Frontier
Poomjai Nacaskul, PhD, DIC, CFA

Machine Learning 4 Banking & Finance (April 1st)
Poomjai Nacaskul
2019.03.31
1. Machine Learning
– Definition & Paradigms
2. Machine Learning
– Algorithms & Languages
3. Machine Learning –
– Application & Frontier
Poomjai Nacaskul
2019.03.31
1. Machine Learning
2. Machine Learning
Poomjai Nacaskul
2019.03.31
Machine Learning – Definition
 What do we mean by Machine?
• Mechanical Calculator
• Programmable Computer
• Environment/State-Aware/Interactive (Physical) Robot
• ARTIFICIAL INTELLIGENCE (AI) powered (Digital-Data) Robot
 What do we mean by Learning?

• Must be based on some kind of Experience, i.e. Data
• Learning from Teacher/Observation/Experimentation
• Ultimate goal is to emulate some form of Intelligence
• Utility regime fast vs. flexible, specificity vs. generality
Poomjai Nacaskul
2019.03.31
Machine Learning – Definition
 Does LINEAR REGRESSION count as machine learning?
• Assume: y =  +  x + (0,2)
• Fit: yi vs. a + b xi , i = 1, … , n
• Minimise: SumSqrdError = i(yi – (a + b xi ))2
• Apply: CALCULUS, i.e. as opposed to ALGORITHM
• Formula: b* = SSxy/SSxx , a* = yaverage – b xaverage ,
where SSxy = i(xi – xaverage)(yi – yaverage), SSxx = i(xi – xaverage)2
 Yes and No!
• Yes, in the sense that a and b does come from 'real-life experience', i.e. in
the form of data set {(xi,yi) , i = 1, … , n}.
• No, in the sense that 'the animal instinctively knew' how to calculate a and
b, i.e. 'intelligence is hard-wired'.
Poomjai Nacaskul
2019.03.31
What is … an Algorithm?
 https://en.wikipedia.org/wiki/Algorithm In mathematics and computer science, …
an unambiguous specification of how to solve a class of problems. … perform
calculation, data processing, automated reasoning, and other tasks.
 https://www.etymonline.com/word/algorithm
1690s, "Arabic system of computation," from French algorithme, refashioned
(under mistaken connection with Greek arithmos "number") from Old French
algorisme "the Arabic numeral system" (13c.), from Medieval Latin algorismus,
a mangled transliteration of Arabic al-Khwarizmi "native of Khwarazm"
(modern Khiva in Uzbekistan), surname of the mathematician whose works
introduced sophisticated mathematics to the West (see algebra). The earlier
form in Middle English wasalgorism (early 13c.),
from Old French. Meaning broadened to any
method of computation; from mid-20c.
especially with reference to computing.
 https://www.youtube.com/watch?v=Ar7CNsJUm58
Poomjai Nacaskul
2019.03.31
Machine Learning – Paradigms
 Learning by Humans
• We learn from Teacher
• We learn from Observation
• We learn from Experimentation
 Learning by Algorithms
• SUPERVISED LEARNING Paradigm
• UNSUPERVISED LEARNING Paradigm
• REINFORCEMENT LEARNING Paradigm
Poomjai Nacaskul
2019.03.31
… but the analogy isn’t perfect …
 Learning by Humans vs. by Algorithms
• Fundamental Difference:
one 'learning architecture’ (i.e. the human brain)
vs. multitudes
• Surprising Difference:
one could be considered a sort of 'black box’,
but which?
• Surprising Similarity:
all boil down to (CONSTRAINED) OPTIMISATION
problem:
min f(x)
s.t. x  X
Poomjai Nacaskul
2019.03.31
Generic Iterative Optimisation Algorithm
 Set up the problem
• Given a problem of the form min f(x) s.t. x  X,
define f, X mathematically & in programming code
 Initialise the solution
• Using some kind of randomisation procedure to generate x0
 Iterative process
• For k = 0 , … ,
check if optimality criterion reached (hence terminate),
otherwise:
find direction of improvementf/x at x = xk ,
determine step size  , then step xk+1 = xk + f/x 
Poomjai Nacaskul
2019.03.31
Generic Supervised Learning Programme
 Modelled Data =
• {(xi,yi), i = 1 , … , n}
 Modelled Mapping =
• x  y = f(x|)
 Modelled Learning =
Iterative Optimisation Algorithm:
0  1      *
that solves, i.e. at algorithmic convergence:
min i(yi – f(xi|))2
s.t.   
Poomjai Nacaskul
2019.03.31
Stochastic Gradient Descent Visualised
 3Blue1Brown (2017), “What is backpropagation really doing?
| Deep learning, chapter 3”, www.youtube.com/watch?v=Ilg3gGewQ5U&t=10m30s
Poomjai Nacaskul
2019.03.31
1. Machine Learning
2. Machine Learning
Poomjai Nacaskul
2019.03.31
Machine Learning – Algorithms
 Supervised Learning – Classifier/Discriminant Analysis
• K-NEAREST NEIGHBOUR (KNN) CLASSIFIER, NAÏVE BAYES CLASSIFIER
• DECISION TREE CLASSIFIER, RANSOM FOREST (RF), GRADIENT BOOSTING (GB)
• LOGISTIC REGRESSION (LOGIT), SUPPORT VECTOR MACHINE (SVM)
• MULTILAYER PERCEPTRON (MLP), FUNCTIONAL-LINK NET (FLN)
• CONVOLUTIONAL NEURAL NETWORK (CNN)
 Supervised Learning – Regression/Time-Series/Seq. Prediction
• RECURRENT NEURAL NETWORK (RNN)
• LONG SHORT-TERM MEMORY (LSTM), Gated Recurrent Units (GRU)
 Unsupervised Learning – Cluster Analysis
 Unsupervised Learning – Dimensionality Reduction
 Reinforcement Learning
Poomjai Nacaskul
2019.03.31
Supervised Learning Fundamentals
 Model Accuracy
• Root Mean Squared Error (RMSE)
• Confusion Matrix, Sensitivity/Recall, Specificity, Precision,
• Receiver Operating Characteristic (ROC) Curve
 Model Generalisability
• Training Set vs. Validation Set vs. Test Set
• Early Stopping, Regularisation
 Model Transparency
Poomjai Nacaskul
2019.03.31
 Model Accuracy
Poomjai Nacaskul
2019.03.31
Confusion Matrix, ROC Curve, etc.
 https://towardsdatascience.com/understanding-confusion-matrix-a9ad42dcfd62
 https://en.wikipedia.org/wiki/Receiver_operating_characteristic
 https://en.wikipedia.org/wiki/Precision_and_recall
Poomjai Nacaskul
2019.03.31
 Model Accuracy
Poomjai Nacaskul
2019.03.31
Early Stopping
 https://deepai.org/machine-learning-glossary-and-terms/early-stopping-machine-learning
 https://www.researchgate.net/figure/The-early-stopping-criterion_fig2_4310358
Poomjai Nacaskul
2019.03.31
Machine Learning – Algorithms
 Supervised Learning – Classifier/Discriminant Analysis
• K-NEAREST NEIGHBOUR (KNN) CLASSIFIER, NAÏVE BAYES CLASSIFIER
• DECISION TREE CLASSIFIER, RANSOM FOREST (RF), GRADIENT BOOSTING (GB)
• LOGISTIC REGRESSION (LOGIT), SUPPORT VECTOR MACHINE (SVM)
• MULTILAYER PERCEPTRON (MLP), FUNCTIONAL-LINK NET (FLN)
• CONVOLUTIONAL NEURAL NETWORK (CNN)
 Supervised Learning – Regression/Time-Series/Seq. Prediction
• RECURRENT NEURAL NETWORK (RNN)
• LONG SHORT-TERM MEMORY (LSTM), Gated Recurrent Units (GRU)
 Unsupervised Learning – Cluster Analysis
 Unsupervised Learning – Dimensionality Reduction
 Reinforcement Learning
Poomjai Nacaskul
2019.03.31
k-Nearest-Neighbour (kNN) Classifier
 http://demonstrations.wolfram.com/KNearestNeighborKNNClassifier
Poomjai Nacaskul
2019.03.31
Decision Tree Classifier
 https://scikit-learn.org/stable/modules/tree.html
Poomjai Nacaskul
2019.03.31
Logistic Regression
 https://en.wikipedia.org/wiki/Logistic_regression
Poomjai Nacaskul
2019.03.31
Logistic Regression
 https://en.wikipedia.org/wiki/Logistic_regression
Poomjai Nacaskul
2019.03.31
Multilayer Perceptron (MLP)
 https://en.wikipedia.org/wiki/Multilayer_perceptron
Poomjai Nacaskul
2019.03.31
What you can do w/ just 2-3 ‘neurons’
angle 1
intercept 1
1 1 1 2 5
temperature 1
5 2
orientation 1 1 1
angle 1
intercept 2
1 1 1 2 5
temperature 2
5 2
orientation 2 1 1
Out[ ]= w eight
What you can do w just 2 artificial neurons

Poomjai Nacaskul
2019.03.31
What you can do w/ just 2-3 ‘neurons’
1st Layer Weights:
ax
Output Neuron on
NeuronA on x,y NeuronA on x,y
bx
NeuronB on x,y NeuronB on x,y
ay
by
az
bz
2nd Layer Weights:

wa
wb Output Neuron on
NeuronA on x,z NeuronA on x,z
Neuron Transfer Functions: NeuronB on x,z NeuronB on x,z
NeuronA LogisticSigmoid Tanh Ramp

Out[ ]=
NeuronB LogisticSigmoid Tanh Ramp
NeuronO LogisticSigmoid Tanh Ramp
Neuron Plot Functions:

PlotNeuronA ContourPlot Plot3D
Output Neuron on
PlotNeuronB ContourPlot Plot3D
NeuronA on y,z NeuronA on y,z
PlotNeuronO ContourPlot Plot3D NeuronB on y,z NeuronB on y,z
Poomjai Nacaskul
2019.03.31
Back Propagation Algorithm Visualised
 3Blue1Brown (2017), “What is backpropagation really doing?
| Deep learning, chapter 3”, https://www.youtube.com/watch?v=Ilg3gGewQ5U&t=4m32s
Poomjai Nacaskul
2019.03.31
How do Neural Networks see the world?
 Nguyen, Yosinski, Clune (2016), “Multifaceted feature visualization:
Uncovering the different types of features learned by each neuron
in deep neural networks”,
http://www.evolvingai.org/files/mfv_icml_workshop_16.pdf
 Two Minute Papers #100
https://www.youtube.com/watch?v=hBobYd8nNtQ
Poomjai Nacaskul
2019.03.31
How Multilayer Perceptron works …
 https://playground.tensorflow.org/#activation=tanh&batchSize=10&dataset=circle&regDataset=reg-
plane&learningRate=0.03&regularizationRate=0&noise=0&networkShape=4,2&seed=0.89941&showTestData=false&discretize=false
&percTrainData=50&x=true&y=true&xTimesY=false&xSquared=false&ySquared=false&cosX=false&sinX=false&cosY=false&sinY=false
&collectStats=false&problem=classification&initZero=false&hideText=false
Poomjai Nacaskul
2019.03.31
… on Nonlinearly Separable Binary Classification Data.
Poomjai Nacaskul
2019.03.31
Much harder to separate 'Double-Coiled' Spiral Data!
Poomjai Nacaskul
2019.03.31
More perceptions would probably help?
Poomjai Nacaskul
2019.03.31
How about stacking them length-wise?
Poomjai Nacaskul
2019.03.31
More efficient to just introduce Functional Links?
Poomjai Nacaskul
2019.03.31
How about Functional Links and more/wider layers?
Poomjai Nacaskul
2019.03.31
When in doubt, go all out?
Poomjai Nacaskul
2019.03.31
Python: scikit-learn – Classifiers
 https://scikit-learn.org/stable/supervised_learning.html
 https://scikit-learn.org/stable/auto_examples/classification/plot_classifier_comparison.html
Poomjai Nacaskul
2019.03.31
Python: TensorFlow 2.0 Alpha
 https://youtu.be/TTQQiJ-mHYA
 https://www.youtube.com/watch?v=k5c-vg4rjBw&t=494s
 https://medium.com/tensorflow/whats-coming-in-tensorflow-2-0-d3663832e9b8
Poomjai Nacaskul
2019.03.31
1. Machine Learning
2. Machine Learning
Poomjai Nacaskul
2019.03.31
Poomjai Nacaskul’s Glossary – p.1/6
ARTIFICIAL INTELLIGENCE (AI) – the notion that human-like intelligence, behaviours and or responses to stimuli
indicative of human intelligence, can be replicated in an artificial apparatus, especially in the form of
programmable digital computer. However, AI is also associated with pre-1980s’ symbolic/semantics-
based EXPERT SYSTEM approach (think of computerised hierarchy of if-then rules used by a medical
doctor to diagnose diseases), hence to be differentiated from ANN-type approach that is inspired by
the computational architecture and signal processing dynamics of biological brains that gained
dominance in 1980’s, the beginning of an era known as “AI Winter”.
BUSINESS INTELLIGENCE (BI) – despite the fact that BI often uses AI, the term intelligence in BI is used in a very
different sense to how it appears in AI. In fact, BI arose as a business-world analog of military
intelligence. As such BI has more to do with the quality of information and the ability to translate
possession of said information immediately into tangible, actionable insights, rather than the ability to
learn from data, form internal model (representation of reality), make associations, infer causalities,
recognise patterns, produce predictions, and/or suggest (approximately optimal) courses of actions,
which are the primary aims of AI research. However, viewed from a software/system engineering
perspective, BI represents an encapsulation/user-interface wrapping of ENTERPRISE ANALYTICS of all kinds,
AI-based or otherwise.
Poomjai Nacaskul
2016.09.06
COMPUTATIONAL INTELLIGENCE (CI) – closer to AI in the use of the term intelligence. Here intelligence is
proxied by the ability to solve problems—finding an answer, finding the best or optimal answer from a
space of all possible answers, finding a means to ensure trajectorial stability vis-à-vis a dynamical
system, and so on. Three strains of research gained prominence: optimisation by means of computer
simulation algorithms inspired by and analogous to the process of genetic-biological evolution by
natural selection (i.e. GENETIC ALGORITHM, GENETIC PROGRAMMING, EVOLUTIONARY ALGORITHM), learning by
means of computational paradigm inspired by and analogous to the human brain (i.e. ARTIFICIAL NEURAL
NETWORK), and mapping between human concepts and machine representation (i.e. FUZZY SETS &
SYSTEMS, HUMAN COMPUTER INTERACTION/INTERFACE).
ENTERPRISE INTELLIGENCE (EI) – used interchangeably with BUSINESS INTELLIGENCE.

MACHINE INTELLIGENCE (MI) – combining the notion that a digital computer can process/possess
knowledge, a la ARTIFICIAL INTELLIGENCE, with the notion that digital devices can be endowed with
sensory organs, thereby able to gather/generate intelligent information about their surroundings and
operating conditions, a la INTERNET OF THINGS. This term is also used to refer to ARTIFICIAL INTELLIGENCE but
without the association with the pre-1980s’ symbolic/semantics-based EXPERT SYSTEM approach, which
itself gets rebranded as GOOD OLD-FASHIONED ARTIFICIAL INTELLIGENCE (GOFAI).
Poomjai Nacaskul
2016.09.06
MACHINE LEARNING – can be thought of as a generalisation of the ARTIFICIAL NEURAL NETWORK paradigm,
incorporating a multitude of learning modalities, knowledge representations, and symbolic/semantic
schemas.
DEEP LEARNING – can be thought of as a modern rebranding of a particular ARTIFICIAL NEURAL NETWORK
architecture dominant in/since 1980’s, the MULTI-LAYER PERCEPTRON (MLP). However, DEEP LEARNING also
distinguishes itself from its 1980’s root by virtue of (i) computational power hitherto unavailable—MLP
with more than a handful of layers of perceptrons used to get bogged down and easily overwhelmed
by large data set, (ii) algorithmic tweaks that prevent connection weights in ARTIFICIAL NEURAL NETWORK
architecture with many interconnected and/or self-connected layers of perceptrons from
degenerating (going to zero or infinity) during training, and (iii) recent breakthrough in
adapting/engineering ARTIFICIAL NEURAL NETWORK architecture for image recognition tasks, notably the
CONVOLUTIONAL NEURAL NETWORK (CNN) and LONG SHORT-TERM MEMORY RECURRENT NEURAL NETWORK (LSTM-
RNN).
QUANTITATIVE MODELS (QM) – the all-encompassing term, covering everything from a basic linear regression
equation, rooting back a few centuries, to the latest DEEP LEARNING innovations of recent months. It is
based on the most undemanding pretext that any entities or phenomena under study can be
indirectly studied/analysed, hence modelled, on basis of their quantitative attributes. As “it’s always
cheaper to play with the toy than to muck about with the real thing”, modelling entities and
phenomena quantitatively saves one from making real mistakes.
Poomjai Nacaskul
2016.09.06
QUANTITATIVE ANALYTICS – the engineering and productionisation of QUANTITATIVE ANALYSIS into readily
deployable tools. The term also refers to a set of (system) statistics themselves, which are produced,
presumably on a regular basis, for the purpose of performance evaluation, fault diagnostics, and
managerial review.
QUANTITATIVE ANALYST – a job description, highlighting the fact that it is the ability to create and work with
QUANTITATIVE ANALYTICS, rather than possession of domain knowledge, that is the critical success factor in
practical applications. A QUANTITATIVE ANALYST may find him/herself in an insurance industry one day,
defense industry the next, and so on. There is a sort of universality with which problems in insurance and
defense industries alike, that can be mapped onto the same/similar mathematical structure, hence
analysed using the same/similar QUANTITATIVE MODELS.
ENTERPRISE ANALYTICS (EA) – refers to the comprehensive set of QUANTITATIVE ANALYTICS used in the strategic,
tactical, as well as routine management of an enterprise, especially large, for-profit corporations, but
also increasingly relevant vis-à-vis small, NETWORK ECONOMY-oriented social enterprises.
Poomjai Nacaskul
2016.09.06
BIG DATA – the term characterising three emergent data trends: (i) sheer volume, (ii) unprecedented
velocity with which bits and pieces of data get generated, pass through digital channels, and arrive at
depositories, and (iii) the variety of format, especially unstructured data streams comprising images,
sound files, and telemetric data logs. BIG DATA also refers to the premise, or rather the promise, that
hidden amongst the volume of high-velocity, multi-variety data lie patterns and nuances that powerful
enough DATA MINING and MACHINE LEARNING algorithms can cipher and distill into useful insights, thereby
uncovering a new set of information that would otherwise not have been captured by traditional
statistics and econometric methods.
DATA MINING – refers to the notion that data is a raw material within which precious, useful bits of
information lie hidden, and also to the process of distilling a vast amount of data for such precious,
useful bits of information, in direct analogy to the process of distilling a vast amount of ores for precious
gems.
DATA SCIENCE – the recognition of availability and arrival of BIG DATA, together with the maturity and utility
of our ability to engineer powerful QUANTITATIVE MODELS to analyse them. Moreover, there is an emergent
area of scientific investigation that study to very dynamics of data objects (be it data bases, network
connectivities, texts and contextual data objects, etc.), how they get generated, evolve,
travel/propagate, and meet their ultimate fate (do bits of data die, or do they live forever?).
Poomjai Nacaskul
2016.09.06
DATA SCIENTIST – either someone who is dedicated to the career of researching DATA SCIENCE proper, or
more commonly a (presumably very good) QUANTITATIVE ANALYST who heads a top-notch QUANTITATIVE
ANALYTICS team.
INTERNET OF THINGS (IOT) – the enhancement of mechanical-electrical devices with ability to gather,
process, generate, store, and report digitally-formatted data and communicate them (nearly) live via
the internet to the intended (sometimes unintended) users.
NETWORK ECONOMY – firstly refers to the extent to which modern economy is driven by and dependent
upon the internet, but latterly refers to the phenomenon by which—as a direct and indirect result of
internet becoming ubiquitous in our daily lives—people, city inhabitants, households, economic
production/consumption units, and so on, begin to interact and relate to one another based on
technology-enabled network topology rather than on the basis of physical proximities, geographic
localities, or even familial ties.
DIGITAL ECONOMY – overlaps with the notion of NETWORK ECONOMY, with perhaps stronger emphasis on the
prevalence of digital devices as key aspects of human productivity, digital communications as key
modes of human connectivity, and digital contents surpassing all other form of archives in human
history.
Poomjai Nacaskul
2016.09.06
bio.
Poomjai Nacaskul received his Bachelor of Arts degree in Physics and Economics, a double major, from WESTERN RESERVE
COLLEGE, CASE WESTERN RESERVE UNIVERSITY (CWRU) in 1992, a Master of Science degree in Operations Research, with Finance
Minor, from CWRU’s WEATHERHEAD SCHOOL OF MANAGEMENT in 1993, and a Doctor of Philosophy in Computational Intelligence
and Operational Research from IMPERIAL COLLEGE LONDON in 1998. Whilst at Imperial, Poomjai also became Quantitative
Analyst with the Quantitative Research and Trading unit of CHEMICAL BANK (UK), where he developed object-oriented
evolutionary optimisation programme for the ‘prop desk’ to explore intraday foreign-exchange trading.
Upon graduation from Imperial College in 1998, Poomjai immediately began his career as an Investment Officer with the
BANK OF THAILAND (BOT)’s INTERNATIONAL RESERVES MANAGEMENT team, later on establishing the bank’s first QUANTITATIVE MODELS
AND FINANCIAL ENGINEERING team, eventually graduating to the Principal Researcher position with BOT’s MONETARY POLICY
GROUP. At BOT, Poomjai researched and developed in cutting-edge areas, from Relative Numeraire Risk in Multi-Currency
Portfolio to Copula Modelling in Pricing Credit Derivatives, and Eigenvector Centrality Analysis of Systemically Important
Financial Institutions.
Poomjai joined SIAM COMMERCIAL BANK (SCB) as Head of the Credit Risk Analytics Department in 2013 and subsequently
created the bank’s first QUANTITATIVE MODELS AND ENTERPRISE ANALYTICS unit, which itself was merged into the bank’s BUSINESS
INTELLIGENCE/DATA ANALYTICS function in 2017, and became one of the bank’s first dedicated Lead/Principal Data Scientist.
At SCB, Poomjai’s involvements reached beyond credit risk, ranging from Predictive Analytics (hourly branch-level
customer queue prediction) to Financial Engineering (validation of Vanna-Volga method for pricing FX exotic derivatives),
and Network Flow Modelling (intra/extra-SCB flows of wholesale liquidity funding).
Poomjai is now a private consultant engaged to UOB ASSET MANAGEMENT (THAILAND). He also teaches Data Analytics and
Machine Learning at CHULALONGKORN UNIVERSITY’s SCHOOL OF INTEGRATED INNOVATION (SCII) and SASIN INSTITUTE OF MANAGEMENT.
Apart from risk management, quantitative finance, machine learning, and data science, Poomjai has keen interest in
Sufficiency Economy Philosophy and Sustainable Development initiatives, as well as AI in medicine/computational biology.
Poomjai Nacaskul
2020.09.30
pub.
• 2019 Talk: 'Fuzzy Multi-Criteria Portfolio Optimisation w/ Python • 2011 Paper (w/ Worawut Sabborriboon): 'Systemic Risk -- Identification,
… and Maybe a Lil' Bit of Mathematica' Assessment and Monitoring based on Eigenvector Centrality Analysis of Thai
Interbank Connectivity Matrices'
• 2019 Lecture: 'Machine Learning Fundamentals'
• 2010 Paper: 'Toward a Framework for Macroprudential Regulation and Supervision
• 2019 Lecture: 'Algorithmic Trading Lectures #2, #1'
of Systemically Important Financial Institutions (SIFI)'
• 2018 Talk: 'Quantum Computing as Disruptive Technology'
• 2010 Lecture: 'Financial Modelling with Copula Functions'
• 2018 Talk: 'Graph-Theoretic Computation in Python'
• 2010 Paper: 'The Global Financial (nee US Subprime Mortgage) Crisis - 12
• 2017 Paper: 'Financial Risk Management and Sustainability The Contemplations from 3 Perspectives'
Sufficiency Economy Philosophy Nexus'
• 2010 Paper: 'Systemic Import Analysis (SIA) – Application of Entropic Eigenvector
• 2016 Publication: 'Survey of credit risk models in relation to capital Centrality (EEC) Criterion for a Priori Ranking of Financial Institutions in Terms of
adequacy framework for financial institutions' Regulatory-Supervisory Concern, with Demonstrations on Stylised Small Network
• 2013 Paper: 'Sufficiency Economy Philosophy: Conceptual Topologies and Connectivity Weights'
Background & Introduction' • 2009 Paper: 'International Reserves Management and Currency Allocation: A New
• 2013 Paper: 'Sustainable Development Calling for Policy Analytics Optimisation Framework Based on a Measure of Relative Numeraire Risk (RNR)'
and (Wittgensteinian Turn Towards) Economics of Moderation' • 2009 Paper (w/ Worawut Sabborriboon): 'Gaussian Slug - Simple Nonlinearity
• 2012 Paper (w/ Kritchaya Janjaroen & Suparit Suwanik): 'Economic Enhancement to the 1-Factor and Gaussian Copula Models in Finance, with
Rationales for Central Banking: Historical Evolution, Policy Space, Parametric Estimation and Goodness-of-Fit Tests on US and Thai Equity Data'
Institutional Integrity, and Paradigm Challenges' • 2001: Paper 'Toward a Framework of Economic Thoughts Based on Sufficiency
• 2011 Chapter: 'Relative Numeraire Risk and the Currency Allocation Economy'
of Sovereign Portfolios' • 1998 Thesis: 'Evolutionary Optimisation and Financial Model-Trading'
Poomjai Nacaskul
2019.10.03
View publication stats
contact
• iPhone  +66(0)90-980-2170
• email  Poomjai.Nacaskul@gmail.com; Poomjai.N@chula.ac.th
• ResearchGate  www.researchgate.net/profile/Poomjai_Nacaskul3
• Linkedin  www.linkedin.com/in/poomjai-nacaskul-phd-dic-cfa-46b994142
• SSRN  http://ssrn.com/author=363974
Poomjai Nacaskul
2020.10.01

Nacaskul Poomjai 2019.03.31.MachineLearningFundamentals - Reformatted

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Nacaskul Poomjai 2019.03.31.MachineLearningFundamentals - Reformatted

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Machine Learning Fundamentals

Presentation · March 2019

Thailand Sustainable Development Foundation View project

Data Analytics & Machine Learning View project

The user has requested enhancement of the downloaded file.

Poomjai Nacaskul, PhD, DIC, CFA

 What do we mean by Learning?

What you can do w just 2 artificial neurons

2nd Layer Weights:

NeuronA LogisticSigmoid Tanh Ramp

NeuronO LogisticSigmoid Tanh Ramp

Neuron Plot Functions:

ENTERPRISE INTELLIGENCE (EI) – used interchangeably with BUSINESS INTELLIGENCE.

• email  Poomjai.Nacaskul@gmail.com; Poomjai.N@chula.ac.th

You might also like