Faculty Digital Archive Collection:
http://hdl.handle.net/2451/14094
Fri, 25 Jul 2014 00:02:52 GMT2014-07-25T00:02:52Z40.72596-73.998345NYU_StatisticsWorkingPapershttp://feedburner.google.comMachine learning for targeted display advertising: Transfer learning in action
http://feedproxy.google.com/~r/NYU_StatisticsWorkingPapers/~3/sa5crxBqzWE/31708
Title: Machine learning for targeted display advertising: Transfer learning in action
Authors: Perlich, C; Dalessandro, B; Stitelman, O; Raeder, T; Provost, F
Abstract: This paper presents a detailed discussion of problem formulation and
data representation issues in the design, deployment, and operation of a
massive-scale machine learning system for targeted display advertising.
Notably, the machine learning system itself is deployed and has been in
continual use for years, for thousands of advertising campaigns (in
contrast to simply having the models from the system be deployed). In
this application, acquiring sufficient data for training from the ideal
sampling distribution is prohibitively expensive. Instead, data are
drawn from surrogate domains and learning tasks, and then transferred
to the target task. We present the design of this multistage transfer
learning system, highlighting the problem formulation aspects. We then
present a detailed experimental evaluation, showing that the different
transfer stages indeed each add value. We next present production
results across a variety of advertising clients from a variety of
industries, illustrating the performance of the system in use. We close
the paper with a collection of lessons learned from the work over half a
decade on this complex, deployed, and broadly used machine learning system.<img src="http://feeds.feedburner.com/~r/NYU_StatisticsWorkingPapers/~4/sa5crxBqzWE" height="1" width="1"/>Tue, 19 Feb 2013 15:41:19 GMThttp://hdl.handle.net/2451/317082013-02-19T15:41:19Zhttp://hdl.handle.net/2451/31708Activism's Impact on Diversified Investors and the Market
http://feedproxy.google.com/~r/NYU_StatisticsWorkingPapers/~3/fcWToiWIPmU/31697
Title: Activism's Impact on Diversified Investors and the Market
Authors: Katz, Barbara; Owen, Joel
Abstract: We model activism as it affects the future distribution of prices in a
portfolio context with risk-averse expected utility of end-of-period
wealth maximizing investors. We characterize activism as affecting the
mean, the variance, and/or the covariance of the target firm’s
price with the prices of the other firms. This characterization allows
us to investigate conditions under which the activist would choose to
become an activist and, subsequently, to derive the sequence of
equilibria that begins with the surreptitious acquisition of shares by
the activist and ends at the moment of the activist’s divestiture
of these shares. We investigate the impact of activism not only on the
target firm’s price over time and the activist’s profit, but
also on the redistribution of portfolio holdings of all market
participants that this activism induces. We propose a method to evaluate
activism and show that, while activism may augment the share price of
the target firm, there exist conditions under which activism would not
necessarily increase the value of the market. Furthermore, we show that
the pro.t of the activist is at the expense of the group of other
investors. We compare our results to recent empirical findings.<img src="http://feeds.feedburner.com/~r/NYU_StatisticsWorkingPapers/~4/fcWToiWIPmU" height="1" width="1"/>Mon, 28 Jan 2013 15:21:28 GMThttp://hdl.handle.net/2451/316972013-01-28T15:21:28Zhttp://hdl.handle.net/2451/31697Drift in Transcation-Level Asset Price Models
http://feedproxy.google.com/~r/NYU_StatisticsWorkingPapers/~3/fldeTR1D_Io/31652
Title: Drift in Transcation-Level Asset Price Models
Authors: Cao, Wen; Hurvich, Clifford; Soulier, Philippe
Abstract: We study the effect of drift in pure-jump transaction-level models for
asset prices in continuous time, driven by point processes. The drift is
assumed to arise from a nonzero mean in the efficient shock series. It
follows that the drift is proportional to the driving point process
itself, i.e. the cumulative number of transactions. This link reveals a
mechanism by which properties of intertrade durations (such as heavy
tails and long memory) can have a strong impact on properties of average
returns, thereby potentially making it extremely difficult to determine
growth rates. We focus on a basic univariate model for log price,
coupled with general assumptions on durations that are satisfied by
several existing flexible models, allowing for both long memory and
heavy tails in durations. Under our pure-jump model, we obtain the
limiting distribution for the suitably normalized log price. This
limiting distribution need not be Gaussian, and may have either finite
variance or infinite variance. We show that the drift can affect not
only the limiting distribution for the normalized log price, but also
the rate in the corresponding normalization. Therefore, the drift (or
equivalently, the properties of durations) affects the rate of
convergence of estimators of the growth rate, and can invalidate
standard hypothesis tests for that growth rate. Our analysis also sheds
some new light on two longstanding debates as to whether stock returns
have long memory or infinite variance.<img src="http://feeds.feedburner.com/~r/NYU_StatisticsWorkingPapers/~4/fldeTR1D_Io" height="1" width="1"/>Tue, 20 Nov 2012 18:02:03 GMThttp://hdl.handle.net/2451/316522012-11-20T18:02:03Zhttp://hdl.handle.net/2451/31652Possible Sharing Arrangements in ARMA Supply Chains
http://feedproxy.google.com/~r/NYU_StatisticsWorkingPapers/~3/mDrT5r2iCGc/31630
Title: Possible Sharing Arrangements in ARMA Supply Chains
Authors: Kovtun, Vladimir; Giloni, Avi; Hurvich, Clifford
Abstract: We introduce a class of new sharing arrangements in a multi-stage supply
chain in which the retailer observes stationary autoregressive moving
average demand with Gaussian white noise (shocks). Similar to previous
research, we assume each supply chain player constructs its best linear
forecast of the leadtime demand and uses it to determine the order
quantity via a periodic review myopic order-up-to policy. We demonstrate
how a typical supply chain player can create a sequence of partial
information shocks (PIS) from its full information shocks FIS and share
these with an adjacent upstream player. We go on to show how such a
sharing arrangement may be benecial to the upstream player by
characterizing the player's FIS in such a case. Hence, we study how a
player can determine its available information under PIS sharing, and
use this information to forecast leadtime demand. We characterize the
value of FIS sharing for a typical supply chain player. Furthermore, we
show conditions under which a player is able to form and share valuable
PIS without (i) revealing its historic demand sequence or (ii) revealing
its FIS sequence. We also provide a way of comparing various PIS sharing
arrangements with each other and with conventional sharing arrangements
involving demand sharing or FIS sharing. We show that demand propagates
through a supply chain where any player may share nothing or a sequence
of PIS shocks with an adjacent upstream player as quasi-ARMA in -
quasi-ARMA out.<img src="http://feeds.feedburner.com/~r/NYU_StatisticsWorkingPapers/~4/mDrT5r2iCGc" height="1" width="1"/>Fri, 05 Oct 2012 15:50:20 GMThttp://hdl.handle.net/2451/316302012-10-05T15:50:20Zhttp://hdl.handle.net/2451/31630