In this post I give an overview of the theory, tools, frameworks and best practices I have found until now around the testing (and debugging) of machine learning applications. I will start by giving an overview of the specificities of testing machine learning applications.
LINK
The aim of this study was to examine whether it is possible to utilize the fluctuations in human motor behaviour to induce a self-organizing process in the athlete, which takes advantage of individual movement and learning characteristics. This recently developed approach is known as differencial learning and is compared to traditional learning. For that purpose, thirty-four recreational skaters participated and practised the speed skating start. A pre- post-test design was used together with a one week intervention period that included three practice sessions of one hour each. The pre- and post-test consisted out of 5 starts, and for each start, the finish time was recorded at a distance of 49 m, which included split time registrations at 5 m, 10 m, and 25 m. Based on the finish time in the pre-test, the participants were equally distributed over three practice groups: a differencial learning, learning by instruction, and control group. Analyses revealed a significant improvement for the differencial learning group in comparison to the control group. It is concluded that differencial learning is an effective method to teach the skating start to novices.
LINK
Athletes who wish to resume high-level activities after an injury to the anterior cruciate ligament (ACL) are often advised to undergo surgical reconstruction. Nevertheless, ACL reconstruction (ACLR) does not equate to normal function of the knee or reduced risk of subsequent injuries. In fact, recent evidence has shown that only around half of post-ACLR patients can expect to return to competitive level of sports. A rising concern is the high rate of second ACL injuries, particularly in young athletes, with up to 20% of those returning to sport in the first year from surgery experiencing a second ACL rupture. Aside from the increased risk of second injury, patients after ACLR have an increased risk of developing early onset of osteoarthritis. Given the recent findings, it is imperative that rehabilitation after ACLR is scrutinized so the second injury preventative strategies can be optimized. Unfortunately, current ACLR rehabilitation programs may not be optimally effective in addressing deficits related to the initial injury and the subsequent surgical intervention. Motor learning to (re-)acquire motor skills and neuroplastic capacities are not sufficiently incorporated during traditional rehabilitation, attesting to the high re-injury rates. The purpose of this article is to present novel clinically integrated motor learning principles to support neuroplasticity that can improve patient functional performance and reduce the risk of second ACL injury. The following key concepts to enhance rehabilitation and prepare the patient for re-integration to sports after an ACL injury that is as safe as possible are presented: (1) external focus of attention, (2) implicit learning, (3) differential learning, (4) self-controlled learning and contextual interference. The novel motor learning principles presented in this manuscript may optimize future rehabilitation programs to reduce second ACL injury risk and early development of osteoarthritis by targeting changes in neural networks.
LINK