Signal to Noise Ratio, flight Performance (Het)

on Monday, October 8th, 2018 2:02 | by

Torque trace flies


Print Friendly, PDF & Email

Bits & Pretzels 2018

The fair took place from 30th September – 2nd Oktober (Sunday to Tuesday). I attended at Monday and Tuesday. The first two days were located at the Messegelände Ost Munich while for the last day the “Schottenhammel” Tent at the Oktoberfest had been reserved.


Visit of the “Corporate Stage” which showcased Interviews:


– The guest came from Adidias but worked independently first- before Adidas she used to make statistical interpretations on her own and present them to companies-> inefficient, it makes more sence to first inquire what the company is interested in- more and more Big Data with high tech factories- example of practical use of statistics: spike in sales as Kanye West fashioned a certain shoe


– generally the new desired type of code is one that can handle uncertanty cases -> away from rule based coding, towards machine learning
– in germany data science is embraced very carefully as no direct value can be assigned to those tasks (in USA quite different)
– at the moment data tends to be scattered, investments have to be made to generate mass data
– should big conservative companies (eg Bosch) accecept help from software companies?
-> going for the long run through data science is a difficult topic in Ger because of the desire fropm investors to generate quick revenue and the strict data safety regulations
Linde Gase
– Linde as a rather conservative company, data science relatively new for them
– Where to go for innovative software design? Silicon Valley?
– Better alternative for Linde: Asia -> more producing Industry, cheaper, still good software developers
The code of the future
– presentation from a guy from stack overflow
– decoding techniques: 1. Rubber duckey debugging: tell a rubber duck on top of your screen what you did, what you want to do and what does not work (93% solved)
2. Divide and conquer debugging: break code up smaller and smaller until you reach the line with the problem
– minority problems in programmer community
– elitary high horse
– importance of algorithms for the public (influence of elections etc.) -> responsability lies with coders
Visit of “main stage”:
Digital revolution – What next?
– big startup investor Albert Wenger
– talk about the sense of income taxes (only ~20% of pupulation paying, majority of taxes comming from a even smaller number)
-> New big economic revolution, capitalism out of date? The battle is more about attention and less about money. Is a universal income necessary with advancing automatisation inevitable?
Emerging trends in Crypto
– How big of a role will Crypto Currency play in the Future?
Digital Competitiveness of nations. The urgency to rule digital innovation.
– Marc Walder (Ringier, Lobbyist for Digitalisation) & Phillip Rösler (former minister of economics in Germany)
– Switzerland having made the transistion to a digitalised country (place 5 on global scale)
– Germany having a long way to go (place 19 on global scale)
– long time consequences can be huge, economics will be more and more dependant on good digital infrastructure
– How to get there? Big companies have to invest and pressure the government (like in Switzerland)
The Innovative Ways Machine Learning is Disrupting Healthcare, Finance and Beyond.
– Tanmay Bakshi (machine learining expert etc.)
– Future of code lies in machine learning
– a lot of advertisement for his finacial management app
Print Friendly, PDF & Email

Program testing & Histogram Comparison

Is the program doing its job? Manual stimulation of the platform:

-> yes it is



Why frequency histograms? I want to compare my data to the data from “Can a fly ride a bicycle?” (Wolf 1992):

example fly E7:


example E10:

 example E11


Histogram of 3 flys (periods of idividual flys have been merged -> one odd and one even period per fly):

Histogram of platform without fly:

Histogram of 11 flys (periods of idividual flys have been merged -> one odd and one even period per fly):

Histogram of 11 flys (no arenaturn):

Fly 3 (discussion model)

All Histograms of different tests for fly 3

-> merging makes little sence if the distribution is completely different areas for each fly!

Print Friendly, PDF & Email

FoxP flight simulator

on Monday, October 1st, 2018 2:55 | by

Print Friendly, PDF & Email

Flight Performance data, Optomotor response

on | by

Print Friendly, PDF & Email

Beta results from first test drive OptoMoto

on Monday, September 24th, 2018 1:56 | by

The different stages of data processing:

1. single fly dyplot with all data present


2. Plot of all the even / odd periods of one fly




3. Plot of all even / odd Periods of one fly merged




4. Plot of the even/odd periods from all tested flys


Another example (first fly tested):



Print Friendly, PDF & Email

Datahandling OptoMotorics

on Monday, September 17th, 2018 12:31 | by

Table with the generated fly test data:

Test Description Average right Average left Differece general
Ex4 normal 4.55382 4.624879 0.071059 4.5893495
Ex4 repositioned 4.562121 4.579682 0.017561 4.5709015
Ex5 normal 4.273299 4.434383 0.161084 4.353841
Ex5 repositioned 4.562121 4.579682 0.1610834 4.5709015
Ex6 normal 3.535669 3.587334 0.051665 3.5615015
Ex7 no arena 3.349097 3.337588 -0.0115097 3.343343
Ex8 bad fly 3.42903 3.394603 -0.034427 3.411816
Ex9 no arena 4.391139 4.363279 -0.0278599 4.377209
Ex9 normal 4.404213 4.456195 0.05198185 4.430204
Ex10 no arena 4.259827 4.195971 -0.0638559 4.227899
Ex10 normal 4.243121 4.360124 0.1170035 4.301622







Print Friendly, PDF & Email

on | by

First Dataset From OptMoto Experiment

on Monday, September 10th, 2018 12:07 | by

Text data file with all the rawdata generated by my new programme: Rawdata OptMoto

R programme which generates the plot:


rawdata <- read.csv(“KHZtext.txt”, header = FALSE, sep = ” “)
flydata <- rawdata[c(3, 1)]
#dygraph(flydata) %>% dyRangeSelector()

arenadata <- rawdata[c(3, 2)]

rawdata <- rawdata[c(3, 1, 2)]

dygraph(rawdata, main = “test graph”) %>%
# dySeries(“V1”, name = “fly”) %>%
# dySeries(“V2”, name = “arena”) %>%
# dyAxis(“y”, label = “Voltage (V)”) %>%

Print Friendly, PDF & Email

reinforcement scores

on Monday, July 23rd, 2018 2:21 | by

Below is given the plot of effect-sizes of reinforcement of 30 genotypes. On the y-axis are the PI values for learning effect sizes.  These scores are calculated by taking the average of PI values of training periods and then subtracting pretest PI values from it.

Reinforcement scores = mean of training score – pretest PI score

Print Friendly, PDF & Email