Professional Documents
Culture Documents
Textbook Intelligent Computing Proceedings of The 2018 Computing Conference Volume 2 Kohei Arai Ebook All Chapter PDF
Textbook Intelligent Computing Proceedings of The 2018 Computing Conference Volume 2 Kohei Arai Ebook All Chapter PDF
https://textbookfull.com/product/intelligent-computing-
proceedings-of-the-2020-computing-conference-volume-2-kohei-arai/
https://textbookfull.com/product/intelligent-computing-
proceedings-of-the-2020-computing-conference-volume-3-kohei-arai/
https://textbookfull.com/product/intelligent-computing-
proceedings-of-the-2020-computing-conference-volume-1-kohei-arai/
https://textbookfull.com/product/proceedings-of-the-future-
technologies-conference-ftc-2018-volume-2-kohei-arai/
Intelligent Systems and Applications: Proceedings of
the 2020 Intelligent Systems Conference (IntelliSys)
Volume 2 Kohei Arai
https://textbookfull.com/product/intelligent-systems-and-
applications-proceedings-of-the-2020-intelligent-systems-
conference-intellisys-volume-2-kohei-arai/
https://textbookfull.com/product/proceedings-of-the-future-
technologies-conference-ftc-2018-volume-1-kohei-arai/
https://textbookfull.com/product/intelligent-systems-and-
applications-proceedings-of-the-2020-intelligent-systems-
conference-intellisys-volume-3-kohei-arai/
https://textbookfull.com/product/proceedings-of-the-future-
technologies-conference-ftc-2020-volume-1-kohei-arai/
https://textbookfull.com/product/advanced-computing-and-
intelligent-engineering-proceedings-of-icacie-2018-volume-2-1st-
edition-bibudhendu-pati-editor/
Advances in Intelligent Systems and Computing 857
Kohei Arai
Supriya Kapoor
Rahul Bhatia Editors
Intelligent
Computing
Proceedings of the 2018 Computing
Conference, Volume 2
Advances in Intelligent Systems and Computing
Volume 857
Series editor
Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland
e-mail: kacprzyk@ibspan.waw.pl
The series “Advances in Intelligent Systems and Computing” contains publications on theory,
applications, and design methods of Intelligent Systems and Intelligent Computing. Virtually all
disciplines such as engineering, natural sciences, computer and information science, ICT, economics,
business, e-commerce, environment, healthcare, life science are covered. The list of topics spans all the
areas of modern intelligent systems and computing such as: computational intelligence, soft computing
including neural networks, fuzzy systems, evolutionary computing and the fusion of these paradigms,
social intelligence, ambient intelligence, computational neuroscience, artificial life, virtual worlds and
society, cognitive science and systems, Perception and Vision, DNA and immune based systems,
self-organizing and adaptive systems, e-Learning and teaching, human-centered and human-centric
computing, recommender systems, intelligent control, robotics and mechatronics including
human-machine teaming, knowledge-based paradigms, learning paradigms, machine ethics, intelligent
data analysis, knowledge management, intelligent agents, intelligent decision making and support,
intelligent network security, trust management, interactive entertainment, Web intelligence and multimedia.
The publications within “Advances in Intelligent Systems and Computing” are primarily proceedings
of important conferences, symposia and congresses. They cover significant recent developments in the
field, both of a foundational and applicable character. An important characteristic feature of the series is
the short publication time and world-wide distribution. This permits a rapid and broad dissemination of
research results.
Advisory Board
Chairman
Nikhil R. Pal, Indian Statistical Institute, Kolkata, India
e-mail: nikhil@isical.ac.in
Members
Rafael Bello Perez, Universidad Central “Marta Abreu” de Las Villas, Santa Clara, Cuba
e-mail: rbellop@uclv.edu.cu
Emilio S. Corchado, University of Salamanca, Salamanca, Spain
e-mail: escorchado@usal.es
Hani Hagras, University of Essex, Colchester, UK
e-mail: hani@essex.ac.uk
László T. Kóczy, Széchenyi István University, Győr, Hungary
e-mail: koczy@sze.hu
Vladik Kreinovich, University of Texas at El Paso, El Paso, USA
e-mail: vladik@utep.edu
Chin-Teng Lin, National Chiao Tung University, Hsinchu, Taiwan
e-mail: ctlin@mail.nctu.edu.tw
Jie Lu, University of Technology, Sydney, Australia
e-mail: Jie.Lu@uts.edu.au
Patricia Melin, Tijuana Institute of Technology, Tijuana, Mexico
e-mail: epmelin@hafsamx.org
Nadia Nedjah, State University of Rio de Janeiro, Rio de Janeiro, Brazil
e-mail: nadia@eng.uerj.br
Ngoc Thanh Nguyen, Wroclaw University of Technology, Wroclaw, Poland
e-mail: Ngoc-Thanh.Nguyen@pwr.edu.pl
Jun Wang, The Chinese University of Hong Kong, Shatin, Hong Kong
e-mail: jwang@mae.cuhk.edu.hk
Rahul Bhatia
Editors
Intelligent Computing
Proceedings of the 2018 Computing
Conference, Volume 2
123
Editors
Kohei Arai Rahul Bhatia
Faculty of Science and Engineering, The Science and Information (SAI)
Department of Information Science Organization
Saga University Bradford, West Yorkshire, UK
Honjo, Saga, Japan
Supriya Kapoor
The Science and Information (SAI)
Organization
Bradford, West Yorkshire, UK
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Contents
v
vi Contents
1 Introduction
Trading is very common in financial industry. Trader will send important information,
as a signal to another party who would be buyer or seller or the IT system or platform.
The question that we are interested in this business scenario is the interactive structure
of a market accounts for informational content, if any, of these potential signals
(Spence 1973) [1], the endogenous market process whereby the trader, for example,
requires (and transmits) information about potential partners (as buyer or seller), which
ultimately determines the implicit speculation involved in selecting portfolios,
allocating information to people and triggering people for further information in the
market? As additional new partners join the market, this study assumes repeated cycles
around the loop to “manage the probable”. This study also modifies trader’s conditional
probabilistic outcomes, as end-reward schedules, are adjusted to partners’ behavior
with respect to signal choice, signal change and post-change to new signals, as they
become available to the trader.
It is useful to introduce a distinction between attributes that collectively constitute
the state of the partner’s present; some are fixed, while others are alterable, but irre-
versible. It is these aspects of the mind or process, as cognitive state, that corresponds
to thinking and feeling, comprising a conglomerate of mental representations and
propositional attitudes (Simon 1973) [2]. At any point in time when interacting with an
individual partner, the trader’s subjective assessment of the speculation which he or she
confronts is defined by these states or conditions, based on probability distributions.
This viewpoint regards signals and states as drivers in probability distributions that
define a trader’s eventual outcome or pay-offs.
At this point it is perhaps clear that there are actions of the trader that “lead the
possibilities”, for a pre-defined outcome or a series of outcomes. These actions, attri-
butes of the dynamics in a complex financial network, are a mathematical function that
takes the trajectory or history of the financial system as an argument and the metrics as
results. As new market information becomes available to the trader through selection
and subsequent observations on partner capabilities, as they represent signals and state,
the trader’s conditional probabilistic outcomes are adjusted, and a new round starts.
The fee schedule of each subsequent new entrant in this network generally differs from
that of the previous partners in the network.
It is desirable to find the expected response, to each action, and impact thereof in
the market over time. It could be something said or written as a reply to a possible
action: something that is expected as a response to an action. To avoid studying the
complex financial system in continuous state of action and response, it is useful to look
at this as a non-transitory configuration of the responsive system, where the market is
generating - either through each “satisficing” decision or a series of small (quantum
like) decisions towards a big decision – an empirical distribution of partner capabilities
through observable action-response or signals (in cognitive states). The trader’s pay-off
or reward, is an intended outcome, of this relationship for information or knowledge,
so yt yp , where the trader’s outcome, yt, and the partner’s outcome, yp, is acceptable
to both party. A reward is as an appetitive stimulus given to a human or some other
entities in recognition of their service, effort or achievement n in order to alter their
behavior. Again, the information or knowledge level conveys the partners’ intent of
investing in information or knowledge in the financial network (Barabasi 2002) [3]. If
they do not invest they might incur lower fees but the loss would exceed the gain from
not making the knowledge-based decisions. This implies that there could be prereq-
uisites that convey no information and hence it is not useful to make any decision based
on that information.
In recent years, due to the advancement in data science and advanced analytics
techniques as well as the massive amount of data available, the vision of better pre-
diction to ‘manage the probable’ and prescription to “lead the possibilities” is gaining
Statistical Learning of Lattice Option Pricing 3
momentum. In the next few sections, we discuss our “slicing” approach of training data
for machines to learn probabilities and their densities: by breaking down the deep
financial network into several layers for a data model, where individual systems –
trader, buyer and seller – interact to take multiple quanta-like small decisions towards a
big decision that naturally generalizes the model (Watts and Strogatz 1998) [4]. We
further slice the data for each machine to learn features and trade-offs, for one asset vis-
à-vis another that likely attracts buyers and sellers to the trader. This multi-layered,
multi-dimensional learning model is designed to determine probabilities and volatilities
– using the Ising-spin model in asymmetric information transition algorithms – for
finding short paths with high probability and low errors for each option. We further use
a multi-layered, multi-dimensional optimization model for efficient learning and
steepest descent to execute the model in a decentralized financial system for dynamic
decisions (Scholtes and Tessone 2011) [5]. Later, we also present some computational
results and performance of the model. Finally, we discuss the limitations of the model
and propose some challenges and opportunities that emerge when using quantum-like
machine learning for strategic decisions with predictable and quantifiable prescriptive
properties.
Individual training data, including asset features, trading (exchange) location, pre-
order, order, delivery, volatility, etc., were structured in a way to provide an under-
standing of the membrane properties of different features of a particular type of partner
or individual. These features of a neuron (partner or individual) in the neural-network
structure show different (labeled) compartments, together with a canonical represen-
tation. Therefore, the search for membrane properties (set to data search), instead of
inferring for deep-layer properties, gives an overview of the membrane properties
found in all compartments in the neuron while connectivity shows the membrane
properties and synaptic connections to all compartments in this neuron and inferences
provides interactions collected at a particular time for all properties of all compartments
in this neuron. The search enables the user to observe the combination of membrane
properties that mediates the integrated activity of a given compartment; compare it to
combined property in different compartments; identify a series of inferences that are
established in a property in a compartment of this neuron type; and learn the differences
using the evidence from a particular property as a signal.
A cognitive and predictive system (Sen 2015) [6] is operable to estimate a human
cognitive state based on activities and trajectories using multiple sources of dynamic
behavioral data and advancing analyses techniques in real-time to draw inferences for
individually-signaled adaptive decisions. Entities may respond to these inferences with
personalized services at a location and at a moment in time that is relevant to the
individual. Inferences for each individual machine may be aggregated to form a col-
lection of dynamic decisions relevant to an individual, and the collection of dynamic
decisions may also be used to draw the inferences for the individual. Representation of
various canonical forms of cognitive states requires the diversity of their dendritic trees,
as in neural-network structure. In a single dendrite represented as an equivalent
4 P. Sen and N. L. Ma
cylinder (e) consists of a chain of three decision compartments, (p) for proximal,
(m) for middle, and (d) for distal, with respect to the cell body. The approach (Shepherd
1968) [7] is representing these patterns as equivalent dendrites, with corresponding p,
m and d compartments.
A lattice fractal is a graph, which corresponds to a fractal, with self-similarities, but
most of them have no translation invariance (Chen 1992) [8]. They are generally
distributed compositional semantics of multi-dimensional objects. Chen’s model shows
that the Ising model on the lattice Sierpinski carpet exhibits the phase transition in any
dimension, but has no transition phase (because of the character of the finitely ramified
fractal) on the lattice Sierpinski gasket. When missing data are present, the model uses
expectation-maximization (EM) algorithm to estimate parameter in Bayesian Networks
when there is training data (Friedman 1997) [9] and in-record linkage when there is no
training data (unsupervised learning). The model further uses EM and Monte Carlo
Markov Chain (MCMC) methods to estimate the error rate automatically in some of the
record linkage situations (Larsen and Rubin 2001) [10].
3 Measuring Trade-offs
Pðk Þ kc
These discoveries have been confirmed in many empirical studies of diverse net-
works. Multi-layered multi-dimensional networks (Newman 2003) [13], as depicted in
Fig. 1, explicitly incorporate multiple service channels of connectivity and constitute
the natural environment to describe systems interconnected through different categories
of connections: each partner channel (relationship, activity, category) is represented by
Statistical Learning of Lattice Option Pricing 5
a layer and the same node or entity may have different kinds of interactions (different
set of neighbors in each layer) for different set of decisions. Partner (dis)satisfaction
measurement efforts generate accurate, actionable insights based on three guiding
principles:
(a) Individuals signal what matter most, without their being asked directly: Explicitly
asking people which they considered as most important to improve in the eco-
nomic delivery, for example, the time required to respond a request versus the
direction and guidance from a staff, is unlikely to produce accurate results.
Unfortunately, most people will say every aspect is equally important. So indi-
vidual machines are trained to learn how to rate each feature (for example, the
overall trading process of settlement) or trade-offs between the features. This
method provides insights about partners’ needs and priorities.
(b) Identify natural break points in service-quality: It is unrealistic to achieve zero
wait times and one-click transactions across all services as it will increase the
operational cost exponentially. Thus, traders need to find a balance between
delivering high-quality and responsive services verses managing resources effi-
ciently and effectively by using metrics to determine “statisficing’ economic
levels. One way for machines to learn is to identify break points – the “last-mile”
journey of experience in trading – at which delays or shortfalls cause (dis)satis-
faction. Machines are able to identify the trade-offs between trader and partner
satisfaction for both of these channels in real time, which will in turn raise overall
partner satisfaction (Faloutsos et al. 1999) [14].
(c) Combine pay-off for internal data with partner data and external public data to
uncover hidden painpoints: Combining trader’s internal information with opera-
tional data of partners – usage data, volume, delay, annual reports, social media
6 P. Sen and N. L. Ma
data, public review, market trends, etc. – can produce additional insights on
machines’ cognitive state implicitly or explicitly. This will help to identify the
correlation between problematic machines and the key drivers of their (dis)sat-
isfaction. The trader system requires a structured model, as in Fig. 2, to foresee
the unintended consequences and “manage the probable” for a collective decision-
making process.
Fig. 2. These steps in decision making give machines the essential elements of a structured
process model.
behavior (e.g. disinvestment). This model also estimates the cognitive state of a
machine, which may include their experience for an activity. A stochastic sub-process
may be executed to determine expectations the probability of a machine (n) being in an
unobserved state r at time t. A state-choice sub-process to determine the cognitive state
for each choice associated with a choice set attribute. The method applied that train one
support vector machine (SVM) per individual choice basket ðb. . .BÞ and to compute
the individual partworths generated from Gaussian distribution by regularizing with the
aggregated partworths. A choice from the choice set (j…J) with feature variables are
shown as X and a choice for one over another is expressed as ð1. . .:iÞ of X, that was
selected by a machine – the trader or partner – is determined and the maximum
likelihood probability for that choice may be used to determine the expected cognitive
state defined as:
1 Y K
Pnr ð jÞ ¼ X ji 1
Bð jÞ i¼1 i
The expected cognitive states for each machine’s decision on a choice-set are
accumulated based on the last activity performed for machines 1 to N. A relational
activity model is used to accumulate the cognitive states. A relational clique is a
construct of a clique over all activities at various states on a trajectory, which may be a
travel path of one or more machines. Each clique C is associated with a potential
function that maps a tuple (values of decisions or aggregations). Together they provide
(a) activity-based decision, (b) state and (c) actions of consecutive activities as
expressed by the following Equation [6].
X Y a
y0 cc
; v0
v0 c c c
c
Next we consider a price model of auctions for an asset in an exchange (stock) market.
Assume that each trader can trade the asset several times at each day t 2 {1, 2, …, T},
but at most one unit number of the asset at each time. Let (t) denote the daily closing
price of tth trading day. And let Kn be a subset of (2), where.
and Ct ð0Þ be a random open cluster on Kn. Suppose that this asset consists of jKn j (n is
large enough) investors, who are located in Kn lattice. And (0) is a random set of the
selected traders who receive the information. At the beginning of trading in each day,
suppose that the investors receive some news. We define a random variable ft for these
investors, suppose that these investors taking buying positions ðft ¼ 1Þ selling positions
ðft ¼ 1Þ, or neutral positions ðft ¼ 0Þ with probability q1 ; q1 or 1 ðq1 þ q1 Þ
ðq1 ; q2 [ 0; q1 þ q2 1Þ, respectively. Then these investors send bullish, bearish or
neutral signal to the market. However, the volatility of the underlying asset is a function of
an exogenous stochastic process, typically assumed to be mean-reverting. Assuming that
only discrete past asset data is available, we adapt an interacting particle stochastic Monte
Carlo algorithm (Moral et al. 2001) [16] to estimate the stochastic volatility (SV), and
construct a multinomial tree which samples volatilities from the SV filter’s empirical
measure approximation at time 0. The study applied the construction of a multivariate
generalization of the Dirichlet-multinomial distribution (DMD), instead of binomial or
trinomial tree, as approximations to continuous-time short-rate models (Kalwani 1980)
[17]. The proposed model of DMD to determine the distribution f(X) of the proportion of
the partners exposed none (F1 = 0), one (F2 = 1), two (F3 = 2)… mn (Fmn+1 = mn) of the
n position of m asset. This model can be expressed in the general form:
Z Z Z Pm2
1 1p0 1 i¼0
pi
f ðXjA; nÞ ¼ ... f ðXjP; nÞDðPjAÞdpm1 . . .. . .dp1 dp0
0 0 0
P
m
with Ai ¼ pi :ðSÞ and S ¼ Ai .
i¼0
The P vector contains the probabilities of m + 1 occurrence (p0 ; p1 . . .pm ¼ P), and
is the multinomial probability vector. The random variable (pi) satisfy the following
constraint:
X
m
pi ¼ 1 and 0 pi 1
i¼0
CðA0 þ A1 þ . . .Am Þ A0 1 A1 1 Am 1
DðPjAÞ ¼ p p1 . . .pm
CðA0 ÞCðA1 Þ. . .CðAm Þ 0
Statistical Learning of Lattice Option Pricing 9
where the A vector {A0, A1, …, Am} is the Dirichlet parameter vector. When the
multivariate Dirichlet and multinomial distribution are compounded, the form of the
proposed model is given as:
P
m Q
m
C Ai Cðxi þ Ai Þ
i¼o
i¼o
Q
m
Cð A Þ C nþ
P
m
Ai
n! i
f ðXjA; nÞ ¼ i¼o i¼o
x0 !x1 !. . .xm !
where X vector {x0, x1…xm} specifies the expected return from each asset of m asset at
n level position opportunities. As with arbitrage-free pricing scheme, we also introduce
one or more Martingale (a.k.a. risk-neutral) probability measures, usually denoted by
Q, which is defined as the SV model with l replaced by r, the risk-free rate. The risk-
neutral price of an option is its discounted expected future value under such a
martingale measure Q. This means that, given the present state of the asset at the time
of pricing, the price must be determined by using the model for the asset under Q.
For simplicity, the study assumes that each partner can be in either one of two spin
states, one state is designated by A, # or +1, and the other by B, " or –1. The total
number of spins (partner states) is B, the number of +1 spins is N or NA and the number
of –1 spin is by B – N. The intensity of action (magnetization), I, may be conveniently
defined as the net number of –1 spins. The intensity of action per spin, I, is then
I N
I¼ ¼ 1 2 ¼ B 2q
B B
where q ¼ NjB. Then the canonical ensemble is:
X
X
Qm ðB; I; T Þ ¼ jm ðT ÞB eEi =kT
i¼1
where Ei is the sum of nearest-neighbor (pair) trade for the ith configuration and jm
represents the non-configurational trade (partition) function of each of the B partners of
the system.
To derive buyer(a)-trader(b)-seller(c) configurations at exchange (d), the system at
exchange (e) limits itself to the simple cubic lattice and allowing nearest-neighbor
interactions that are not only between habihbcihcd i and hdei, but also hbd i (trader and
exchange) in a three-dimension Ising model as nearest-neighbor interactions in a self-
similar way does affect critical behavior (Gefen et al. 1980) [18]. We derive the model as:
ð1 þ xÞ3 ð1 þ yÞ X ur ð yÞur
Kðy; xÞ ¼ :
4 2 r 0 ð1 þ yÞ2r
10 P. Sen and N. L. Ma
Fig. 3. Features and attributes of assets at different buying and selling process.
PG ðn pÞ ¼ pm :pnðn1Þ=2m
This simple stochastic model for networks has been used in the modeling of a
variety of structural features of real-world networks. This mapping predicts that the
common epithets used to characterize competitive systems, such as “winner takes all,”
“fit get rich” (FGR), or “first-mover advantage,” emerge naturally as thermodynami-
cally and topologically distinct phases of the underlying complex evolving network
Statistical Learning of Lattice Option Pricing 11
(Bianconi and Barabasi 2001) [19]. To demonstrate the existence of a state transition
from the FGR phase, the pay-off distribution follows:
gðÞ ¼ C 2h
where h is a free parameter and the pay-offs are chosen from 2 ð0; max Þ; with
þ 1
normalization C ¼ h þ 1= hmax : For this class of distributions the cognitive state (a
Bose condensation) is:
hþ1 Z
b max xh
dx \1
ðb max Þh þ 1 b min ðtÞ ex 1
Fig. 4. The same mechanisms of learning models that lead to the improvements in transitions
with larger lattices for bother partners (M) and Traders (E) – a myopia of tendency to ignore the
larger picture.
there exists a relationship between trader and partners in the financial network (in terms
of its eigenvalues) and the dynamic processes are happening at the level of nodes.
Dynamic replication processes also take place in a discrete space. (Farbod 2009)
[22] uses the discrete logarithmic mapped in a graph for such process, it is equivalent to
stochastic search, and transport phenomena within the network. This study introduces
the basic idea of the replica method by computing the spectrum of random matrices. To
be confident, the study explores the case of real symmetric matrices of large order
N having random elements Jij that are independently distributed with zero mean and
variance J 2 ¼ N 1 .
The resolvent R ^ of the N N matrix ^J is defined as follows:
^ ðeÞ ¼ eI ^j 1
R
where I is the N N identity matrix and e is a complex number. This also formulates
^ ðeÞ:
in computing the trace of R
X 1
^ ðeÞ ¼
Tr R
k
e ek
Statistical Learning of Lattice Option Pricing 13
in the large N limit. One can argue that, decision-making is not invariant to the physical
environment in which a decision is made. In addition, a partner with access to deeper
(quantum) information resources may be able to do strictly better than a partner with
access only to classical information resources.
In this respect, our findings are somewhat akin to those in computer science that
have established the superiority of quantum over classical algorithms for certain
problems. In the nutshell the function R^ ðeÞ is a real analytic function with a cut on the
real line and the discontinuity on the cut is the density of states. Now we can write the
average of the logarithm of the determinant in the following way (Dotsenko 2005) [23]:
n=2
d 1
^
log det eI j ¼ 2 lim
n!0 dn detðeI ^
j
Operatively one computes the quantity on the r.h.s for integer (positive) values of
the replica number n, then one performs an analytical continuation of the result to real
values of n and eventually takes the limit n ! 0. Such replication in decision-making
in complex systems is based on multi-layered, multi-dimensional-optimization problem
with several criteria to identify the sensitive preferences of a decision maker.
The key idea of quantum decisions (QD), decentralized deeper (quantum) signals in
the decision-making, as, is to provide the simplest generalization of the underlying
individualized multiple decisions towards a single decision, so as to account for the
complex dynamics of the many non-local hidden variables that may be involved in the
cognitive trade-offs and decision making processes of the brain (Yukalov and Sornette
2009) [24]. In such decision-making, involving signals with unknown states of nature,
actions of trade-offs, responses to actions, and rewards in subconscious play the role of
hidden variables. However, the learning model represented the neural network as a
quantum-like object for which several mechanisms have been suggested (Lockwood
1989) [25]. It also meant a genuine quantum nature of some elements in the psycho-
logical processes in decision-making (Cheon and Takahashi 2010) [26]. However, to
rigorously test the existence or the absence of genuinely quantum effects, the study, in
Fig. 5, considered a decision-making experiment with incomplete information, analo-
gous to the Harsanyi type quantum game (Cheon and Iqbal 2008) [27] in which
inequalities (Bell 1964) [28] were tested. In the case of uncertainty, when the outcomes
are unknown, partners may lack a clear reason for choosing an option and consequently
they could abstain and make an irrational choice. When partners confronting uncer-
tainty — paralyzing them against acting — are presented with a detailed explanation of
the possible outcomes, they changed their mind and decided to act, thus reducing the
disjunction effect (Tversky and Shaffir 1992) [29].
Thus, by providing the partner with additional explanations, it is possible to
influence their state or conditions. This gives the partner a drive to either alter their
existing cognitions, or to reframe their interpretation of a situation, through a re-
orientation of their local frame. Rather than positing a collapse of the partner’s state to
whichever axis represents their decision, this model updates their decision by shifting it
towards the axis representing the decision.
14 P. Sen and N. L. Ma
Fig. 5. Results show big fluctuations near the critical point as the behavior of (Partner) M vs
(Trader) E is very different.
where ti – set of time period {1, 2, …, T} at stage i; Dit – total partners buy-sell at stage
0 00
i at time t; Dit þ c – total partners buy-sell at stage i0 at time t + c; Dit þ k – total partners
0
buy-sell at stage i00 at time t + k; ait - buy-sell rate at stage i at time t; ait þ c – buy-sell rate
00
at stage i0 at time t + c; ait þ k – buy-sell rate at stage i00 at time t + k; lit – trade rate at
0 00
stage i at time t; lit þ c – trade rate at stage i0 at time t + c; lit þ k – trade rate at stage i00 at
0
time t + k; Cti – delay cost at stage i at time t; Cti þ c – delay cost at stage i0 at time t + c;
00
Cti þ k – delay cost at stage i00 at time t + k; Rt – Available trades at time t; R0t þ c –
Available trades at time t + c; R00t þ k – Available trades at time t + k. The model
includes
1 if trades are assigned at stage i at time t
Xti ¼
0 otherwise
Another random document with
no related content on Scribd:
intervenção, pedida no ultimo trimestre, já parece o caso decidido
que veiu a ser no meiado de 47.
O pobre Espectro, estonteado, inconsequente, enrouquece para
proferir palavras de paz. Tambem elle teme a victoria dos seus:
«Para que é incitar o povo a que entre no palacio dos nossos reis e
ahi pratique acções de cannibaes? O paço dos nossos reis é um
foco de corrupção politica, mas não o é de corrupção moral. Não ha
rainha mais virtuosa como esposa, nem como mãe de familias». (n.
27) O Espectro, cabisbaixo, com a voz que fôra eloquente, parece
pedir perdão do clamor que levantára outr’ora, exigindo a abdicação
d’essa propria rainha e a limpeza da caverna de Caco. Porque é
uma tal mudança? Foi o inglez (Seymour) que veiu (fevereiro) a vêr
se conseguia compôr os partidos em armas, a vêr se evitava ainda a
acção da Hespanha; e o Espectro, e a gente que escolhera para seu
orgão a voz de uma sombra, na esperança do poder, moderavam a
furia, mudavam o rumo. Mas como rainha e governo se recusaram a
pactuar com a revolta (abril), a voz do Espectro voltou a vilipendiar
aquella que tres mezes antes era o modelo das virtudes:
A côrte pela sua parte, toda sybarita, toda gastronomica,
entra ainda na lucta com intenção doble. Se vence, o
systema absoluto triumpha; se succumbe, acceita as
condições e entrega os cadaveres dos seus amigos em
holocausto á nacionalidade offendida, á moral publica
ultrajada ...
E quando, afinal, a Inglaterra teve de annuir cabalmente ás
exigencias da França-Hespanha, consequindo apenas tomar
tambem parte na intervenção; quando as forças extrangeiras
chegaram, para que os chefes da revolta, não sabendo que fazer
d’ella, se lhe entregassem, fingindo obedecer á força, o Espectro,
voltando a achar a eloquencia dos primeiros dias, dizia nos ultimos:
A côrte, o ministerio, o rei, tudo isso desappareceu. Não
caíram ás nossa mãos, que nol-as ataram; mas sumiram-se
na voragem de um protocollo. Isso que ahi se chama rei é um
espantalho, os ministros são os lacaios de lord Palmerston.
Foi a rainha, foram os Cabraes, quem nos vendeu, quem nos
trahiu ... (n. 63. julho, 3)
Não seria acaso tambem o conde das Antas, indo metter-se com
toda a sua gente na bocca do lobo inglez que viam aberta á barra do
Douro? Não seria tambem Sá-da-Bandeira, por não querer passar
de Setubal a Lisboa? Não seriam tambem os chefes da revolução
armada, politicos, generaes, cortezãos em vez de tribunos, com
medo da demagogia que passaria por sobre elles, se acaso
consummassem a victoria? A intervenção servir-lhes-hia, se a
Inglaterra a fizesse a favor do seu partido: como a fez a favor do
inimigo, a intervenção era um crime.
Quanto este espectro é com effeito a imagem sumida da viva
personalidade do tribuno Passos em Belem! Dez annos bastaram
para mumificar a democracia; n’esses dez annos, os seus chefes
tinham fechado o Arsenal, dissolvido os batalhões, entregando-se
nos braços da ordem. Dez annos (32-42) tinham bastado tambem
para que o desenvolvimento necessario das premissas postas na
legislação liberal apparecesse: a liberdade era um absolutismo da
nova aristocracia dos ricos nascida da concorrencia; e em vez de
Mousinho, um Bentham, apparecia imperando um homem duro e
pratico, o conde de Thomar, imagem ossea de um systema já
consolidado.
Democratas, liberaes, eram agora todos, as sombras dos que
tinham sido. A miseria crua do paiz reduzia-os á condição de seres
famintos, amesquinhando-lhes os caracteres, baixando-lhes a
estatura, avolumando só a podridão natural das covas. Que o
extrangeiro viesse a este cemiterio afastado, mandar a cada
espectro para a sua tumba, acabar a funebre revolta de cadaveres,
não admirava ninguem, porque a liberdade trouxera-a elle. Estava
obrigando a manter o systema. Mas, dez annos havia, o extrangeiro
encontrara cá um povo singular, extravagante, um dos sete
dormentes da Europa, inaccessivel ás idéas novas, mas vivo,
abraçado de joelhos ao throno-altar. Agora, voltando, o extrangeiro
só via tambem o espectro d’esse povo antigo: sombras errantes
falando uma linguagem archaica tremida nos labios brancos de
frades rotos e senectos; cordões de mulheres luctuosas ajoelhadas
perante os nichos allumiados, resando «por alma dos nossos irmãos
que foram mortos n’esta rua!»
Uns dos mortos voltavam para sempre aos eternos jazigos; outros
fugiam do velho cemiterio das doutrinas, deitavam fóra o lençol da
liberdade, e a correr batendo os ossos, vestiam as fardas
regeneradoras lantejouladas, e, mirando-se em trajos de vivos,
ficavam crendo ter resuscitado.[32]
4.—A PRIMAVERA DE 47
NOTAS DE RODAPÉ: